Termodinamica Ocean

ARTICLE IN PRESS
Journal of Theoretical Biology 249 (2007) 124–139

www.elsevier.com/locate/yjtbi
A thermodynamic perspective on food webs: Quantifying entropy

production within detrital-based ecosystems
Filip J.R. Meysmana,, Stijn Bruersb
a
Netherlands Institute of Ecology (NIOO-KNAW), Centre for Estuarine and Marine Ecology, Korringaweg 7, 4401 NT Yerseke, The Netherlands
b
Instituut voor Theoretische Fysica, Katholieke Universiteit Leuven, Celestijnenlaan 200D, B-3001, Leuven, Belgium
Received 5 April 2007; received in revised form 17 July 2007; accepted 17 July 2007
Available online 20 July 2007
Abstract
Because ecosystems fit so nicely the framework of a ‘‘dissipative system’’, a better integration of thermodynamic and ecological
perspectives could benefit the quantitative analysis of ecosystems. One obstacle is that traditional food web models are solely based upon
the principles of mass and energy conservation, while the theory of non-equilibrium thermodynamics principally focuses on the concept
of entropy. To properly cast classical food web models within a thermodynamic framework, one requires a proper quantification of the
entropy production that accompanies resource processing of the food web. Here we present such a procedure, which emphasizes a
rigorous definition of thermodynamic concepts (e.g. thermodynamic gradient, disequilibrium distance, entropy production, physical
environment) and their correct translation into ecological terms. Our analysis provides a generic way to assess the thermodynamic
operation of a food web: all information on resource processing is condensed into a single resource processing constant. By varying this
constant, one can investigate the range of possible food web behavior within a given fixed physical environment. To illustrate the
concepts and methods, we apply our analysis to a very simple example ecosystem: the detrital-based food web of marine sediments. We
examine whether entropy production maximization has any ecological relevance in terms of food web functioning.
r 2007 Elsevier Ltd. All rights reserved.
Keywords: Non-equilibrium thermodynamics; Ecology; Modeling; Dissipation; Ecosystem
1. Introduction Thermodynamics offers a different, complementary per-

spective on food web functioning. Thermodynamic ap-
The biological components within ecosystems engage in proaches focus on the various process rates induced by the
a complex process of self-organization so that food webs food web, and this requires a detailed description of the
emerge (Solé and Bascompte, 2006). A fundamental communication with the ‘‘outside’’ environment.
question in ecology is whether these food webs are shaped To examine whether thermodynamic constraints are
by (1) the interplay of internal interactions, or (2) external relevant for food web operation, one clearly needs a proper
thermodynamic constraints, or (3) a combination of both. thermodynamic description of food webs. This marriage
Up to now, the ecological literature has primarily explored between ecology and thermodynamics continues to be a
the first option, looking at architecture, topology, complex- true challenge. Although the need for a thermodynamic
ity etc.—see reviews in de Ruiter et al. (2005); Montoya et perspective has long been recognized (Schroedinger, 1944),
al. (2006). Roughly speaking, one could say that the these two disciplines have shown little integration over the
ecological perspective primarily emphasizes the ‘‘inside’’ of past 50 years. This dichotomy between ecology and
the food web (network organization of the food web, the thermodynamics is nicely illustrated by the distinct model
role of trophic interactions), and because of this, the currencies that are in use within different research
abiotic environment receives a very crude description. communities. In the field of ecological modeling, the
default state variables are mass and energy, of which the
Corresponding author. Tel.: +31 113 577 489; fax: +31 113 573 616. associated conservation equations form the central heart of
E-mail address: f.meysman@nioo.knaw.nl (F.J.R. Meysman). food web and ecosystem models (Lindeman, 1942; May,
0022-5193/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jtbi.2007.07.015
ARTICLE IN PRESS
F.J.R. Meysman, S. Bruers / Journal of Theoretical Biology 249 (2007) 124–139 125
1981; Grover, 1997; Loreau and Holt, 2004). The theory of within an ecological context. This communication has
non-equilibrium thermodynamics basically extends this therefore two central goals. The first goal is to illustrate
mass/energy toolbox with the entropy, the central concept how this theory can be applied in a strict ‘‘orthodox’’
to the Second Law of thermodynamics. A principal focus fashion to the analysis of resource processing within
in thermodynamic analysis is therefore the proper quanti- ecosystems (i.e. without invoking any ‘‘ecological’’ addi-
fication of entropy production within various model tions). To this end, we need to correctly translate
systems (Glansdorff and Prigogine, 1971; Nicolis and thermodynamic concepts (e.g. dissipation, thermodynamic
Prigogine, 1977; Kondepudi and Prigogine, 1998). None- gradient, distance from equilibrium) into an ecological
theless, the term ‘‘entropy’’ does usually not feature in context. As noted above, entropy is the central variable in
ecological textbooks (e.g. May, 1981; Begon et al., 1996; the analysis of dissipative systems, and hence, a prime
Case, 2000). The application of non-equilibrium thermo- concern will be to develop a consistent procedure to
dynamics in biology is classically restricted to the calculate the entropy production within an ecosystem. The
subcellular level, where metabolic processes are cast within second goal is to examine whether the entropy production
the formalism of chemical thermodynamics (Sandler and rate has any ecological relevance in terms of food web
Orbey, 1991; Kurzynski, 2006). Yet, when moving to the functioning. In recent years, a number of propositions have
community or ecosystem level, these concepts and methods been made along this line, in which the notion of ‘maximal
are no longer invoked. The concept of entropy remains entropy production’ surfaces in relation to ecosystem
highly enigmatic to most ecological modelers, while operation (Ulanowicz and Hannon, 1987; Schneider and
thermodynamicists have not applied their entropy produc- Kay, 1994; Svirezhev, 2000; Fath et al., 2001; Schneider
tion calculations to actual ecosystems. and Sagan, 2005). Between these studies, there is however
In recent years, there are a growing number of considerable difference in the actual implementation of this
publications that specifically aim to close the gap between entropy production maximalization, and its subsequent
ecology and thermodynamics (see reviews in Jorgensen, interpretation. Moreover, the treatments often remain
2000; Jorgensen and Svirezhev, 2004; Schneider and Sagan, conceptual and vague, and propositions have not been
2005). Given its central role in physical applications of convincingly tested against actual data. To examine
non-equilibrium thermodynamics, one would expect the whether and how entropy production is relevant to food
entropy concept to play a dominant role in this emerging webs, we will perform a case study of the detrital-based
field of ‘‘thermodynamic ecology’’. With some notable ecosystem of the ocean floor. This ecosystem has the
exceptions (Aoki, 1995; Svirezhev, 2000; Aoki, 2006), this advantages that (1) it is chemotrophic, and so it can be
is surprisingly not the case. Instead, the prevalent idea is fully described with the standard chemical thermodynamics
that the standard mass/energy/entropy toolbox of non- (and hence we do not need to revert to radiation
equilibrium thermodynamics is not sufficient to describe thermodynamics as in phototrophic systems), (2) suitable
ecological processes (Jorgensen, 2000; Jorgensen and data are available from past biogeochemical studies to test
Svirezhev, 2004). In this view, ecology is inherently more thermodynamic constraints, and (3) the procedure can be
complex than the dissipative phenomena usually consid- generalized to other chemotrophic ecosystems (such as
ered in ‘‘abiotic’’ non-equilibrium thermodynamics, so that chemo-autotrophic vent systems and terrestrial soils).
additional constructs and new rules must be introduced to
fully describe an ecosystem’s functioning. As a result, a 2. An ecosystem model of the ocean floor
number of novel ‘‘ecological’’ concepts have been for-
warded as essential complements to the standard entropy 2.1. Mass balance formulation
concept, including emergy (Odum, 1983), exergy (Schnei-
der and Kay, 1994), eco-exergy (Susani et al., 2006), and The ocean floor covers nearly 75% of the earth surface,
ascendancy (Ulanowicz and Abarca-Arenas, 1997). How- and apart from some notable but localized exceptions
ever, the use of these newly defined properties does not (thermal vent systems, coral reefs, seagrass beds, coastal
come without problems. The first problem is the loose and phototrophic mats), the bulk of the ocean floor harbors a
verbal use of thermodynamic language, and the highly detrital-based, chemo-heterotrophic ecosystem (Burdige,
qualitative way in which these concepts are used (e.g. see 2006). This system is fueled by organic carbon that is fixed
discussions in Mansson and McGlade, 1993; De Wit, by photosynthesis within the upper ocean, settles through
2005). Secondly, each of these concepts has its proponents, the water column, and rains down on the sediment surface.
but the exact interrelation between them remains unclear. Within the sediment, part of this organic matter input is
Thirdly, and more fundamentally, the question remains taken up by the food web, and converted into biomass and
whether the standard mass/energy/entropy toolbox of non- metabolic end products. A small fraction ð5%Þ of the
equilibrium thermodynamics is really insufficient to ana- incoming flux escapes degradation and is buried into
lyze ecosystem functioning? deeper layers (Middelburg and Meysman, 2007).
Despite some pioneering work (e.g. Aoki, 1995, 2001; To arrive at a suitable model description of the carbon
Svirezhev, 2000), believe that the theory of non-equilibrium dynamics within the ocean floor ecosystem, the conceptual
thermodynamics has not been explored to its full potential scheme of Fig. 1a is used. This scheme reduces sedimentary
ARTICLE IN PRESS
126 F.J.R. Meysman, S. Bruers / Journal of Theoretical Biology 249 (2007) 124–139
l l
l
Fig. 1. (a) Idealized flow scheme of the detrital-based ecosystem of marine sediments. The ecosystem features an arbitrary complex food web, which is
treated as a black box. The direction of the arrows specifies the convention by which flows are taken positive. (b) Specific example of the general scheme
depicted on the left: In this basic food web, one consumer is feeding on a single resource. See text for an explanation of the different flows.
organic matter processing to its barest essentials. The the ecosystem exchanges matter: one resource reservoir at
principal goal is to illustrate the quantification of entropy concentration C 0R , and one waste reservoir at concentration
production within ecosystems, rather than providing a C 0W . For simplicity, we assume that the exchange of
detailed model of ocean floor biogeochemistry. Our model organic matter takes place with a single reservoir (instead
universe contains two compartments, termed ‘ecosystem’ of incorporating the buried organic matter into a separate
and ‘environment’. The ecosystem refers to the bioturbated external reservoir). This broad-brush ‘pooling’ of all
zone of the ocean floor, which is a relatively thin sediment external organic matter suitably simplifies the derivations,
layer of thickness L (typically around 10–30 cm), char- but does not fundamentally change the thermodynamic
acterized by biological sediment reworking and intensive analysis or the conclusions arrived upon.
processing of organic matter (Boudreau, 1998; Middelburg The principal difference between the internal ecosystem
and Meysman, 2007). This intense biogeochemical activity reservoirs and those of the environment is their size. The
results from an interplay between microbial decomposition external reservoirs are considered ‘infinite’, so that the
and the reworking activity by sediment dwelling inverte- concentrations C 0R and C 0W are fixed model parameters.
brates (Kristensen et al., 2005). All resident organisms Consequently, the operation of the ecosystem is con-
(viruses, microbes, meiofauna, invertebrates) are collec- strained by a fixed physical environment. In contrast, the
tively referred to as the food web, which is treated as a internal ecosystem reservoirs have a finite size, and are
black box in Fig. 1a. The environment refers to all external governed by following mass balances
surroundings, which includes both the overlying water
dC R
as well as the deeper lying sediment. Temperature effects ¼ F R F RB , ð1Þ
are ignored, and so both ecosystem and environment dt
remain at the same temperature T. In our model analysis, dC W
¼ F W þ F BW , ð2Þ
we assume that the operation of the food web is only dt
limited by the availability of organic food resources. In dC B
¼ F RB F BW , ð3Þ
other words, organic matter processing is not constrained dt
by the availability of electron acceptors used in respiration, where t represents time, and the flows F i are expressed as
which is a typical assumption in biogeochemical model mass per unit of volume and time. The flow F RB represents
studies of marine sediments (e.g. Berg et al., 2003; the net uptake of resource by the food web, while F BW is
Meysman et al., 2003). As a result, we do not need to the net release of CO2 by the food web. The constitutive
consider the mass balance dynamics of the electron expressions for these flows will depend on the structure of
acceptors explicitly (e.g. O2 , SO4 ). Instead, we only track the resident food web, and so one needs an explicit food
the behavior of two carbon compounds, termed resource web representation (as shown below). The flows F R and
(R) and waste (W). These can be respectively identified as F W , respectively, denote the net exchange of resource and
the organic matter food substrate, represented by the waste with the environment, and are parameterized as
simplified stoichiometry CH2 OjR , and CO2 , the metabolic
end-product of respiration. F R ¼ aR ðC 0R C R Þ, ð4Þ
The model ecosystem consists of three carbon reservoirs: FW ¼ aW ðC 0W C W Þ. ð5Þ
one resource reservoir, one waste reservoir, and one
reservoir that pools all biomass of the food web (at This linearized transfer is the standard way of modeling the
respective concentrations C R , C W and C B ). The environ- exchange with the environment in ecological models
ment is abstracted as two separate reservoirs with which (Tilman, 1982; Loreau and Holt, 2004). We assume that
ARTICLE IN PRESS
the transport coefficients aR and aW are not dependent of metabolism. Yet, for the present purpose, it is sufficient,
upon food web activity. Although it is known that fauna as it incorporates those features that are important from a
can significantly influence the exchange between sediment thermodynamic perspective. Basal maintenance refers to
and external environment (Meysman et al., 2006), such energetic costs to ‘stay alive’, and is described by the
feedbacks are not considered here. Accordingly, the transformation
coefficients aR and aW can be directly calculated from the
dominant physical transport processes that are operating in CH2 OjB þ O2 ! CO2 þ H2 O, (7)
the sediment. The flow F R represents the difference where the associated rate F M refers to maintenance
between the deposition flux of organic matter F D , and respiration. The turn-over of biomass due to mortality
the burial flux F B (Boudreau, 1997; Burdige, 2006). The (from predation, accident, disease, ageing) is described by
deposition flux can be written as F D ¼ ðfo=LÞC 0R , with f
the porosity, o the burial velocity sediment and C 0R the CH2 OjB ! CH2 OjR , (8)
organic carbon content of the incoming material. The
with associated turnover rate F T . Biomass turn-over thus
export to deeper layers can be parameterized as
means that biomass simply disassembles into basic building
F B ¼ ðfo=LÞC R , which now features the organic carbon
blocks, which become again available for biomass synthesis
concentration within the sediment. The exchange of CO2
or respiration. Together this implies that F RB ¼ F A F T
with the overlying water is parameterized by the diffusive
and F BW ¼ F M þ ð1 qÞF A , and so the mass balances for
exchange F W ¼ ð3fD=L2 ÞðC 0W C W Þ, where D is the pore
this very simple one-consumer food web become
water diffusion coefficient of CO2 . Accordingly, one finds
that aR ¼ fo=L and aW ¼ 3fD=L2 . dC R
¼ FR FA þ FT, ð9Þ
dt
2.2. Example: a simple single-consumer food web dC W
¼ F W þ F M þ ð1 qÞF A , ð10Þ
dt
In the above model formulation of the ocean floor dC B
¼ qF A F M F T . ð11Þ
ecosystem, the food web structure was treated as a black dt
box, and the mass balances (1)–(3) hold for any (arbitrarily In the next sections, we will introduce the necessary
complex) food web. This black box perspective will be a theoretical tools to calculate the entropy production
cornerstone of our thermodynamic analysis. However, as associated with this particular food web. This will confront
an example illustration, we now analyze an explicit us with the yet unresolved conceptual problem of assigning
representation of the food web. To this end, we consider chemical potentials to organisms.
the simplest possible ecosystem: only a single consumer is
feeding upon the available resource R. This is not meant as
a realistic representation of the actual food web in marine 2.3. Quantifying entropy production
sediments, and in fact, it hardly deserves the name food
web, as only a single organism is present. Still, because of In the standard formulation of non-equilibrium thermo-
its simplicity, it didactically illustrates concepts. For dynamics, the entropy production si associated with a
reference, one could think of the microbial community in given flow F i is calculated as the product of that flow with
marine sediments that is treated as a single population. Fig. a corresponding thermodynamical force X i (Nicolis and
1b shows the transformations. We assume that the Prigogine, 1977; Kondepudi and Prigogine, 1998)
consumer’s biomass CH2 OjB is simply assembled from
elementary building blocks of resource CH2 OjR . The si ¼ F i X i . (12)
synthesis of biomass does not occur spontaneously, and
Accordingly, we need to calculate the proper forces X i that
must always be coupled to respiration. This coupling
are associated with each of the five flows F i in our example
between anabolism and catabolism can be formally
ecosystem model (Eqs. (9)–(11)). The resource and waste
represented via the chemical transformation
exchange essentially describe a mixing process between two
CH2 OjR þ ð1 qÞO2 reservoirs A and B at different concentrations. The
! qCH2 OjB þ ð1 qÞCO2 þ ð1 qÞH2 O. ð6Þ associated thermodynamic force can be directly calcu-
lated from the difference in the chemical potential m
The associated reaction rate F A represents the assimilation of these reservoirs via the relation X mix ¼ DG mix =T ¼
rate. In reaction (6), O2 generically represents the electron ðmA mB Þ=T, where DG mix is the Gibbs free energy of
acceptor, while CO2 and H2 O stand for the end-products mixing and T is temperature (Kondepudi and Prigogine,
of respiration. The yield factor q specifies the amount of 1998). If we treat resource and waste as ideal solutes,
biomass that results from the assimilation of one unit of their chemical potential scales with the logarithm of
resource. Or equally, 1 q denote the respiration costs the concentration, i.e. m ¼ mref þ RT lnðC=C ref Þ, where R
associated with biomass synthesis (activity respiration). denotes the universal gas constant, and the concen-
Obviously, this reaction provides a very simplified picture tration C ref refers to some reference state. As a result,
ARTICLE IN PRESS
we obtain 2.5. Entropy balance formulation

X R ¼ ðm0R mR Þ=T ¼ R lnðC 0R =C R Þ, ð13Þ The Second Law of Thermodynamics requires that for
XW ¼ ðm0W mW Þ=T ¼ R lnðC 0W =C W Þ. ð14Þ each independent flow F i the associated entropy produc-
tion si should be positive (Nicolis and Prigogine, 1977;
Accordingly, the associated entropy production rates
Kondepudi and Prigogine, 1998). A given flow is ‘inde-
become
pendent’ when it is not coupled to any of the other flows. In
sR ¼ F R R lnðC 0R =C R Þ, ð15Þ our model, all five flows are independent, and so the
associated entropy production rates should all be positive.
sW ¼ F W R lnðC 0W =C W Þ. ð16Þ
Note that a proper linkage of anabolism to catabolism
The biotic flows F A , F M and F T are in essence rates through the yield factor q must ensure that sA becomes
associated with a given chemical reaction. The thermo- positive.
dynamic force associated with a chemical reaction can be Accordingly, we also find that sEM 40 and stot 40
calculated from the relation X reac ¼ DG reac =T, where because their individual components are all positive. To
DG reac denotes the Gibbs free energy of the reaction. provide further insight into the issue of entropy generation,
This quantity is calculated from the difference in chemi- we can write an entropy balance for the model universe as a
cal potential between reaction products and reactants whole. We can always decompose the total entropy of the
(Kondepudi and Prigogine, 1998), and so we obtain universe S uni into separate contributions of ecosystem
(‘‘sys’’) and environment (‘‘env’’), Suni ¼ S env þ Ssys . The
X A ¼ ðð1 qÞmW þ qmB mR Þ=T, ð17Þ resulting entropy balance thus becomes
X M ¼ ðmW mB Þ=T, ð18Þ
dSuni dSenv dS sys
X T ¼ ðmR mB Þ=T, ð19Þ ¼ þ ¼ stot . (22)
dt dt dt
where mB , mR and mW , respectively, denote the chemical Because the model universe is isolated as a whole, no
potential of biomass, resource and waste. Note that the entropy export occurs across its boundaries, the entropy
assimilation force simply forms a linear combination of change is only due to internal entropy generation stot .
the other two forces, i.e., X A ¼ ð1 qÞX M X T . When the Moreover, because stot 40, Eq. (22) assures that the
flows F A , F M and F T are known, one can multiply them entropy of our model universe as a whole can only
with the force expressions (17)–(19) to calculate the increase, in line with the classical Second Law statement
associated entropy production rates. Together, the five for isolated systems.
entropy production rates sR , sW , sA , sM and sT fully
quantify all ‘dissipation’ that occurs within our model uni- 2.6. The chemical potential of biomass?
verse, where the term dissipation is thus simply synon-
ymous to entropy production. Although theoretically valid, the entropy production
expressions (17)–(19) require an explicit expression for the
2.4. Two entropy production rates chemical potential of the biomass mB . Here we are
confronted with one of the real challenges in thermo-
Two composite rates of entropy production can be dynamic ecology. To our knowledge, the issue of how to
derived that are of particular interest. Firstly, the calculate the chemical potential for a population of living
‘ecosystem entropy production’ accounts for the entropy organisms has not yet been properly resolved. In general,
production that occurs within the ecosystem alone, i.e., the the chemical potential can be written as mi ¼ hi Tsi where
entropy production exclusively associated with the func- hi is the specific enthalpy and si is the specific entropy
tioning of the food web. Summation of the entropy (where ‘specific’ means per mole or kg of substance).
production rates associated with the biotic flows results in Theoretically at least, the specific enthalpy of an organism
can be determined by standard calorimetric methods (by
sEM sA þ sM þ sT . (20)
putting a suitable number of the organisms in a bomb
where the subscript ‘EM’ refers to ecosystem metabolism. calorimeter, and measuring the resulting heat of combus-
Secondly, we can calculate the entropy production that tion—although this may cause practical/ethical problems).
occurs within the complete model universe (both ecosystem Yet, no theoretical procedure exists to calculate the specific
and environment). Summation of all five entropy produc- entropy of an organism at a given population density.
tion rates leads to Three basic obstacles prevent this. Firstly, there is the
question whether it is possible at all to attribute a well-
stot sR þ sW þ sA þ sM þ sT ¼ sR þ sW þ sEM . (21)
defined entropy to a complex, highly-organized, far-from-
This ‘total entropy production’ encompasses the dissipa- equilibrium entity such as an organism? Some authors
tion that occurs within the ecosystem ðsEM Þ, as well as the argue that this is possible (Battley, 1999; Aoki, 2001), but
dissipation that results from the communication with the the debate has not been convincingly settled. Assuming
external surroundings (sR and sW ). that it is possible, a second obstacle is that biomass
ARTICLE IN PRESS
comprises a complex mixture of different compounds, interactions can be highly non-linear. The only assumption
which also comes in ‘biomass packages’ of different size we make is that for some fixed boundary conditions C 0R
(i.e. organisms). Bottom-up approaches have been pro- and C 0W , the ecosystem will eventually reach a steady state,
posed that start from the basic entropy properties of the thus making abstraction of the possibility of oscillatory
different molecules (lipids, sugars, proteins), to calculate and chaotic dynamics. For simplicity, we also assume that
the entropy change that corresponds to the formation of a single stable state is reached, thus ignoring the possibility
1 g of biomass starting from a dilute aqueous solution of of multiple stable states. To make a distinction with
inorganic nutrients (Morowitz, 1968; Battley et al., 1997). transient properties, steady state values are denoted by an
However, such procedures are typically based on intracel- asterisk superscript . Because we treat the food web as a
lular conditions in micro-organisms, and whether they black box, we cannot say anything about its internal state
apply to macro-organisms such as rabbits or tigers remains (e.g. the distribution of biomass over various trophic
to be proven. More importantly, bottom-up procedures levels). Yet, we do know that in a steady state, the inputs
seem to ignore that organisms come in different sizes and and outputs must cancel each other, while the total food
form populations of different densities. To see this, web biomass must remain constant dC B =dt ¼ 0. Conse-
consider two hypothetical prey populations, i.e. a zebra quently, Eq. (3) directly reveals that
population at a certain biomass density (kg biomass per
F RB ¼ F BW . (23)
m2 ), and a rabbit population at exactly the same biomass
density. In a first approximation, one can assume that In other words, biomass synthesis should match the sum of
zebras and rabbits consist of roughly the same type of maintenance and turn-over, and the net metabolic activity
biomass (composition in terms of lipids, sugars, proteins). of the food web only consists of respiration
If so, then the bottom-up procedure would predict about CH2 OjR þ O2 ! CO2 þ H2 O. (24)
the same specific entropy for both species. Thermodyna-
mically, there would be no distinction between zebra and Effectively, this reaction symbolizes the core dissipative
rabbit populations, although the allocation of organic process within our ecosystem: high-quality resources
matter in space is very different. Effectively, the organic (CH2 O, O2 ) are converted into low-grade waste products
matter in the zebra population is ‘concentrated’ into in a (CO2 , H2 O). The associated reaction rate is referred to as
few large packages, while the organic matter in the rabbit the ‘‘ecosystem metabolism’’ F EM . Imposing the steady-
population is distributed as many small packages. We state constraint onto the resource and waste balances (1)
believe this crucial difference should also be reflected in the and (2), we find that the ecosystem metabolism must equal
specific entropy. the other flows
F EM ¼ F RB ¼ F BW ¼ F R ¼ F W . (25)
3. Steady state analysis
Because mR ¼ mref
þ RTR lnðC R =C ref
and mW ¼R Þ mref
W þ
The above analysis of the simple single-consumer model RT lnðC W =C ref
W Þ, the associated thermodynamic force of
has exposed the key problem with entropy production in the respiration reaction (24) becomes
food webs. The mass balances (1)–(3), the entropy X EM ¼ ðmW mR Þ=T ¼ R lnðK eq C R =C W Þ, (26)
expressions (12)–(21), and the entropy balance (22) are
valid both in steady as well as transient regimes. where K eq ¼ expððmref
R mref
is the equilibrium
W Þ=ðRTÞÞ
Unfortunately, the inability to calculate the chemical constant of the respiration reaction. For a given reference
potential of biomass mB prevents the use of these state, this quantity is a constant. Combining the above
expressions to obtain the entropy production associated relations with expressions (20) and (21), we obtain
with food web operation. Here, we propose a way to following steady-state entropy production rates
circumvent this difficulty. This is done by only considering sEM ¼ F EM R lnðK eq C R =C W Þ, ð27Þ
the steady-state, which refers to the situation where the
stot ¼ F EM R lnðK eq C 0R =C 0W Þ. ð28Þ
properties of the ecosystem no longer vary with time. In the
steady-state, the mB quantities will drop from the equations These expressions do no longer contain the quantity mB ,
(as shown below). This ‘‘trick’’ forms the cornerstone of and so, we have circumvented the previously discussed
our thermodynamic analysis of food webs. This trick not difficulty of defining the chemical potential of the biomass.
only applies to the specific food web of the previous The total entropy production depends on the external
section, but is generally valid for any food web. boundary conditions imposed upon the ecosystem (via C 0W
and C 0R ) and on the internal structure of the food web (via
3.1. Steady state mass and entropy balance F EM ). In contrast, the ecosystem entropy production
depends on the internal ‘abiotic’ state of the ecosystem
Consider an arbitrarily complex food web, whose (via C W and C R ) and on the internal ‘biotic’ structure of
functioning is governed by some internal dynamics. We the food web (via F EM ).
do not put any constraint on this internal complexity: Note that the term ‘‘steady-state’’ should be interpreted
many trophic compartments may be present and the with caution: the time invariance only applies to the
ARTICLE IN PRESS
ecosystem, and not the environment. In the non-trivial A more meaningful interpretation of the distance from
situation where there is a food web operating, and so equilibrium is possible when looking at the disequilibrium
F EM 40, the environmental reservoirs do change their mass between resource and waste products. For the chemo-
(the resource reservoir loses mass at rate F EM , the external trophic ecosystems studied here, the term ‘thermodynamic
waste reservoir gains mass at rate F EM ). The assumption equilibrium’ refers to the chemical equilibrium of the
that C 0W and C 0R are fixed is nothing but a suitable respiration reaction (28). Two situations can be distin-
approximation for large reservoirs with a slow relaxation guished, depending on whether thermodynamic equili-
time. Also, when applying the steady-state condition to the brium applies to the environment or to the ecosystem
entropy balance (22), only the term dS sys =dt related to the
Dmenv ¼ m0R m0W ¼ RT lnðK eq C 0R =C 0W Þ, ð31Þ
ecosystem vanishes, and so one obtains
Dmsys ¼ mR mW ¼ RT lnðK eq C R =C W Þ. ð32Þ
dS env
¼ stot 40. (29) Note that both these distances are defined in terms of
dt
abiotic state variables (i.e. resource and waste concentra-
This illustrates that in the ‘‘steady-state’’, the properties of tions), and that they are expressed in units of energy. For
the environment should not remain constant in time. convenience, we can also define corresponding quantities
Effectively, the Second Law requires that stot 40, and so that are expressed in units of mass (by linearly expanding
the entropy balance (29) necessitates that the entropy of the the logarithm in (31) and (32)). The mass-based analogue
environment should always increase. In other words, when of Dmenv thus becomes
the ecosystem resides within a steady-state, all the entropy
that is generated (i.e. within the ecosystem or through C 0W
D ¼ C 0R . (33)
interactions between ecosystem and environment), will K eq
ultimately ‘accumulate’ within the environment.
This quantity expresses how far the external environment is
from thermodynamic equilibrium. In the vocabulary of
3.2. Ecosystems as far-from-equilibrium entities non-equilibrium thermodynamics, D is referred to as the
‘‘thermodynamic gradient’’ that is imposed as a boundary
One frequently encountered quote in texts on ecological condition upon the (eco)system. Obviously, the environ-
thermodynamics is that ‘‘ecosystems operate far-from- ment is in thermodynamic equilibrium when D ¼ 0, and in
equilibrium’’. However, a rigorous quantitative interpreta- this ‘‘ultimate state’’, no resource conversion can take place
tion of this ‘‘distance from equilibrium’’ is typically not and no food web can exist. In a similar fashion, we can
given. In the classical interpretation, the far-from-equili- introduce the mass-based analogue of Dmsys as
brium connotation refers to the thermodynamically ‘im- C W
probable’ state of living organisms (Jorgensen and d ¼ C R . (34)
K eq
Svirezhev, 2004). In this view, organisms represent a highly
self-organized form of matter, which resides in a low This quantity expresses the internal disequilibrium, or how
entropy state as compared to their surroundings. This far the ecosystem operates from thermodynamic equili-
intuitive idea can be formally expressed as brium in steady-state. The ecosystem is in thermodynamic
equilibrium when d ¼ 0. The physical interpretation of D
Dm ¼ mB mref , (30) and d is further discussed below.
where mref denotes some reference potential, usually
introduced as ‘‘the chemical potential that biomass would 3.3. The ecosystem resource processing regime
have if it were brought in equilibrium with its surround-
ings’’ (e.g. Jorgensen, 2000). Although intuitively appeal- Intuitively, it is clear that—at least in theory—many
ing, there are two major problems with this definition. different food webs can operate within a given physical
First, as explained above, a major roadblock in thermo- environment. However, to test this statement in a true
dynamic ecology is the impossibility to calculate the quantitative sense, we need a rigorous mathematical
chemical potential mB of an organism. This renders the infilling of both the terms ‘‘food web operation’’ and
definition (30) of little use in actual calculations. Second, ‘‘physical environment’’. This can now be done by the tools
and more fundamentally, there is also a problem with the introduced in the previous sections. In our model universe,
reference state. In real world situations, neither the the specification of the ‘‘physical environment’’ comes
ecosystem, nor the environment can reside in a state of down to the specification of a set of five abiotic parameter
thermodynamic equilibrium (see below). As a consequence, values: the concentrations in the external resource reser-
it is not possible to ‘‘bring biomass in equilibrium with its voirs C 0R and C 0W (and so the thermodynamic gradient
surroundings’’, because the surroundings are not in D40 is fixed), the transport coefficients aR and aW , and the
thermodynamic equilibrium themselves. Consequently, equilibrium constant K eq .
there is no ‘‘natural’’ equilibrium state that can serve as a Now assume that (1) the physical environment remains
logical reference to define mref . constant in time, (2) a certain (arbitrary complex) food web
ARTICLE IN PRESS
has established itself within this environment, and (3) this identifier of a given food web. Each food web will show a
food web operates in a steady state. This state will always certain value for F EM or kEM . Yet, different food webs can
be characterized by a particular value for the ecosystem show the same value of F EM or kEM , and hence, different
metabolism F EM . Now suppose one knows F EM (e.g. by food webs can correspond to the same ERPR.
measuring CO2 accumulation in a closed incubation
system). This sole quantity is now sufficient to completely
3.4. The ERPR of the single-consumer food web
characterize the resource processing by the food web,
provided the external environment is fully characterized
The calculation of kEM is functional, as it condenses all
(C 0R , C 0W , aR , aW , and K eq are known). Indeed, one can
information on the resource processing of a given food
directly calculate the associated resource and waste
web. This enables a straightforward comparison of the
concentrations from the mass balances (1)–(2) as
operation of different food webs (see below). Yet, it is
C R ¼ C 0R F EM =aR , ð35Þ crucial to understand that kEM is not a true ‘‘kinetic
constant’’, as in general, it shows a complex dependency on
C W ¼ C 0W þ F EM =aW . ð36Þ
the (non-linear) internal dynamics of the food web. To
From these concentrations, the disequilibrium distance d, illustrate this, we can revisit the simple food web above
the flows (F EM , F R and F W ) and the entropy production with one consumer. To use the mass balances (9)–(11), we
rates (sEM , sR and sW ) can be subsequently calculated need explicit rate laws for the biotic flows F A , F M and F T .
using the expressions of the previous sections. We will refer The assimilation rate is given a Holling type I response
to this particular set of variables as the ‘‘ecosystem F A ¼ rA ðC R C W =K eq ÞC B , where rA is the assimilation
resource processing regime’’ (ERPR). The ERPR comple- rate parameter. This is a classical ecological rate law but
tely describes the overall effect of the food web on resource for the factor C W =K eq , which is added for thermodynamic
processing, but does not specify any ‘‘internal’’ food web consistency. Free energy constraints require that the
characteristics (biomasses, trophic flows). In essence, the synthesis of new biomass should stop when the ecosystem
ERPR adopts a black box approach to the resource is at chemical equilibrium, i.e., when C R ¼ C W =K eq .
processing of food webs. Maintenance respiration and biomass turnover are simply
A crucial question is then how fast resources are taken linear in the biomass, i.e., F M ¼ d M C B and
converted? By combining the quantities F EM , C R and F T ¼ d T C B , where d M and d T are rate constants. The
C W , we can define the resource conversion rate parameter of values of C R , C W and C B then can be determined from the
the ecosystem as steady state versions of the mass balances (9)–(11).
A bifurcation occurs when the thermodynamic gradient
1 C W 1 exceeds the threshold value Dc ¼ ðd T þ d M Þ=ðqrA Þ. This
kEM ¼ F EM d ¼ F EM C R : ð37Þ
K eq critical threshold is only dependent on the biotic para-
The inverse of kEM specifies the time scale over which meters of the food web (here, the consumer parameters q,
resources are degraded within the ecosystem. Expression rA , d M and d T ). Effectively, this point serves as ‘threshold
(37) can be used to calculate kEM from F EM , but one can for life’. When the thermodynamic gradient is too small
also work the other way round. To show this, we first ðDoDc Þ, one obtains the trivial solution C R ¼ C 0R ,
introduce the auxiliary rate constant C W ¼ C 0W , and C B ¼ 0. In other words, no stable
consumer population can be established, and no resource
1
1 1 conversion occurs within the ecosystem. Beyond the
kc ¼ þ : ð38Þ threshold value ðD4Dc Þ, the food web starts operating,
aR aW K eq
and the ecosystem properties take the values
The parameter kc defines the characteristic time scale at
kc
which material is exchanged between ecosystem and C R ¼ C 0R ðD Dc Þ, ð40Þ
environment. The value of kc will be dominated by the aR
kc
rate limiting transfer (note the analogy with the formula for C W ¼ C 0W þ ðD Dc Þ, ð41Þ
two serial resistances in an electrical network). When aW
qkc
aR 5K eq aW one obtains kc aR . Upon substitution of the C B ¼ ðD Dc Þ. ð42Þ
concentration expressions (35) and (36) into (37), one ð1 qÞd T þ d M
directly finds If one substitutes the expressions for C R and C W into the
kEM definition (34) of the internal distance from equilibrium,
F EM ¼ D. (39) one finds d ¼ Dc . Accordingly, the internal distance from
1 þ kEM =kc
equilibrium is independent from the external thermody-
Accordingly, there are two equivalent ways to fully namic gradient, and is solely dependent on the properties
characterize the ERPR: either by specifying F EM or by of the consumer. The ecosystem metabolism in steady state
fixing kEM . Both quantities serve as master parameters that becomes
suitably summarize the resource processing associated with
a particular food web. Note that the ERPR is not a unique F EM ¼ ð1 qÞF A þ F M ¼ kc ðD Dc Þ ¼ kEM Dc . (43)
ARTICLE IN PRESS
From this, the resource conversion parameter can be intense organic matter processing, (2) Young Sound (YS),
calculated as an arctic site in Northeast Greenland, and (3) the Santa
Barbara Basin (SBB), an anoxic basin off the coast of
D
kEM ¼ kc 1 . (44) California. These three sites are selected because they
Dc
are very different in terms of sediment type, geochemistry,
This shows how kEM is ultimately dependent on both the and resident food web. A second reason is that their
physical environment (through kc and D) and the internal biogeochemistry has been recently studied in great
details of the food web (via Dc ). The external environment detail (AB: Fossing et al., 2004; YS: Berg et al., 2003,
is characterized by a parameter set (C 0R , C 0W , aR , aW , and SBB: Meysman et al., 2003). Due to these detailed
K eq ). Similarly, the food web is characterized by a para- biogeochemical studies, the physical environment is well
meter set (q, rA , d M and d T ). Both parameter sets together characterized, which allows us to directly calculate the
thus determine the value of kEM . Accordingly, this simple parameters of the physical environment in our model (C 0R ,
example clearly shows that kEM , and hence the ERPR, is C 0W , aR and aW ).
intrinsically dependent on the food web structure. Moreover, these studies provide detailed quantitative
This simple food web also illustrates the connection information on in situ organic processing. For our purpose
between the two disequilibrium distances d and D here, it is particularly relevant that reliable values for the
introduced above. In economic terms, the thermodynamic deposition flux F D and the resource processing rate kdata
gradient can be seen as a measure of potential activity. The have been estimated. The biogeochemical studies refer-
conversion of resources into waste products can only take enced above distinguish between three types of organic
place when the thermodynamic gradient is sufficiently matter, and hence report three separate values of F D and
large, i.e. when D4Dc . When D is large, there a large kdata . (1) A fast decaying fraction, which is rapidly
potential for resource transformation within the ecosystem degraded and disappears within the top mm of the
(good ‘business opportunities’ for the consumer). When D sediment; (2) A slow decaying fraction, which disappears
becomes small, the opportunities to benefit from the within the top 20 cm; (3) Refractory material, which only
dissipation reaction (24) become smaller and smaller (poor shows reactivity over geological time scale (this material
‘business climate’). Whether any resource processing only decays on a scale of meters to hundredths of meters of
actually occurs, and how fast it proceeds, depends on the sediment depth). As our study concerns organic matter
details of the food web. Hence the use of the term processing in the bioturbated zone (10–20 cm), we selected
‘potential’. The external thermodynamic gradient effec- the slowly decaying fraction as the relevant type of organic
tively constrains the potential range of the internal matter for our analysis.
disequilibrium distance, i.e., 0pdpD. The disequilibrium A compilation of parameters is given in Table 1. All but
distance d that is effectively realized depends on the two parameters are constrained by measurements at all
particular food web structure (in our case on the consumer three sites, and parameter values are directly taken from
properties q, rA , d M and d T ). The food web essentially the references. This availability of sufficient parameter
determines how fast resources are processed as compared data was exactly the reason why the three sites were
to transport processes that supply resources and remove selected. No data are available for the thickness L of the
waste products. When resources are rapidly processed, the sediment, and the equilibrium constant of respiration K eq .
ecosystem is poised close to thermodynamic equilibrium Accordingly, these latter two parameters were given ‘best
(kEM ! 1 and d ! 0). In contrast, when resources are estimates’, and afterwards, an uncertainty analysis was
slowly processed, the internal disequilibrium approaches performed to check how the results were dependent on
the maximal value (kEM ! 0 and d ! D). Note than when these estimates. To constrain the sediment thickness, we
D ! 0 one automatically obtains d ! 0: when the used for all three sites the global average thickness of the
environment resides in thermodynamic equilibrium, one bioturbated zone, i.e., L ¼ 10 cm (Boudreau, 1998). The
automatically obtains that the ecosystem also must be in value of the equilibrium constant K eq is difficult to
thermodynamic equilibrium (in this trivial situation noth- precisely constrain under natural conditions. However,
ing happens at all). we know that it is very high for the aerobic degradation of
organic matter (reaction 24). Under standard thermody-
4. Thermodynamic analysis of ecosystems namic conditions (298.15 K, 1 atm pressure, 1 molar
concentrations), and using glucose as the organic substrate,
4.1. A comparison of three marine sediment ecosystems one finds a standard free energy of reaction DG0 ¼ 532 kJ
per mole of carbon (Canfield et al., 2005), leading to a
Equipped with the thermodynamic concepts and model- value of K eq ¼ expðDG 0 =RTÞ 1093 . In actual sediment
ing tools introduced in the previous sections, we can now environments, organic matter will consist of a mix of
take up the actual thermodynamic analysis of ecosystems. substrates that release less free energy than glucose, and so
To illustrate concepts and techniques, we will examine K eq will be lower. Still, the value of K eq will remain high,
three different marine sediment ecosystems: (1) Aarhus Bay and we employed K eq ¼ 108 for all three sites as an initial
(AB), a temperate coastal site in Denmark characterized by estimate.
ARTICLE IN PRESS
Table 1
Parameter set for three marine sediment ecosystems: Aarhus Bay (AB) in Denmark, Young Sound (YS), and Santa Barbara Basin (SBB)
Input data AB YS SBB
Burial velocity o cm yr1 0.32 0.12 0.26

Porosity f 0.80 0.65 0.87
Temperature T 1C 10 0 6
Diffusion coefficient CO2 D cm2 263 160 270
Deposition flux carbon FD mol m2 yr1 5.3 1.8 0.6
CO2 bottom water C 0W mol m3 2.5 2.2 2.45
Sediment thickness L cm 10 10 10
Model parameters
Resource exchange aR yr1 25 103 8 103 23 103
Waste exchange aW yr1 6.3 3.1 7.0
External resource C 0R mol m3 2084 2265 263
External waste C 0W mol m3 2.5 2.2 2.45
Equilibrium constant K eq – 108 108 108
Resource conversion rates
Critical value kc yr1 0.026 0.008 0.023
Reported value kdata yr1 0.379 0.095 0.056
MEP value kmep yr1 0.509 0.159 0.442
Resource conversion efficiency Z % 94 93 72
The input data are taken from detailed biogeochemical studies at these sites: AB (Fossing et al., 2004), YS (Berg et al., 2003), and SBB (Meysman et al.,
2003). The model parameters are directly calculated from these input data based on the formulas in the text. The data value kdata is the resource conversion
parameter as reported in the above references. The characteristic value kc is calculated from the model parameters through Eq. (38). The value of kmep is
found by implementing the constraint of maximal entropy production Eq. (45).
As detailed above, each food web will be associated with exchange with the environment. In this scenario, the ERPR
a certain ecosystem resource processing regime (ERPR). is said to be under ‘‘reaction control’’ and the associated
The ERPR describes the operation of the food web in a food web is categorized as ‘‘slow’’. Expression (39) reveals
black box fashion, and is completely determined when the that F EM kEM D, and so the ecosystem metabolism
value of resource processing parameter kEM is specified. linearly scales with the resource processing parameter
However, within the same physical environment an infinite (note that the semi-logarithmic plotting in Fig. 2a obscures
number of ERPR are possible. In theory, the value of kEM this linearity). As expected, under reaction control, the
can range from zero to infinity, and the ERPR of the resource piles up within the ecosystem, while waste levels
resident food web will lie somewhere in between these two are low (Fig. 2b). Oppositely, when kEM bkc , the internal
endpoints. In the next sections, we will proceed in two resource conversion proceeds much faster than the physical
steps. Initially, we make no assumption on the feasibility of exchange with the environment. In this scenario, the ERPR
a certain ERPR for each of the three sites examined. So in is said to be under ‘‘transport control’’, and the associated
a first step, we will only use the available data on the food web is categorized as ‘‘fast’’. When the resource
physical environment of the sites. For each site, we will processing rate parameter increases, the ecosystem meta-
scan the whole range 0okEM o1, and for each kEM value, bolism rapidly approaches a constant value F EM kc D
we will calculate the corresponding ERPR. In a second (Fig. 2a). Under transport control, the resource is depleted
step, we will examine where the actually observed ERPR by the high food processing activity within the food web,
(i.e. the ERPR corresponding to kdata ) lies on this scale while waste products accumulate (Fig. 2b).
0okEM o1. When taken to the extreme, the slow and fast food webs
give rise to two end-member situations. The ‘‘minimal’’
4.2. Slow and fast food webs end-member is attained when there is no biological activity
at all, and so kEM ! 0, we find that F EM ! 0. In this limit
As noted above, the rate constant kc provides a also, the resource and waste concentrations will simply
characteristic time scale for the physical transport between match those of the external reservoirs, i.e. C R ! C 0R and
ecosystem and environment. It also serves as a threshold to C W ! C 0W (Fig. 2b), and so, the distance from equilibrium
distinguish between two broad classes of ERPR, which we becomes maximal and equals the thermodynamic gradient,
refer to as fast and slow food webs. Fig. 2 illustrates this for i.e. d ! D. Oppositely, the ‘‘maximal’’ end-member is
the Aarhus Bay site (the two other sites show a similar attained when biological activity is very high: resources are
pattern). When kEM 5kc , the internal resource conversion immediately processed by the food web the moment they
within the ecosystem is slow compared to physical become available, and so kEM ! 1. Note however that
ARTICLE IN PRESS
4000 12
Ecosystem metabolism (Aarhus Bay) Concentrations (Aarhus Bay)
6
Ecosystem metabolism FEM (mol m-2 yr-1)
Resource concentration (mol m-3)
Waste concentration (mol m-3)

5 Fmax 3000 9
EM
slow food web fast food web

4 (reaction control) (transport control)
2000 *
6
3 slow food web fast food web Resource (CR)
(reaction control) (transport control)
2
1000 3
1 Waste (C*W)
κcrit κdata
κcrit κdata
0 0 0
10-4 10-2 100 102 10-4 10-2 100 102
Resource processing rate κEM (yr-1) Resource processing rate κEM (yr-1)
Fig. 2. (a) The ecosystem metabolism F EM plotted as a function of the specific metabolic rate kEM for the Aarhus Bay site. (b) The resource C R and waste
C W concentrations plotted as a function of the specific metabolic rate kEM for the Aarhus Bay site. In each panel the actually observed ERPR is plotted as
kdata .
the associated ecosystem metabolism still reaches a finite archaea, viruses). Organic processing in SBB is dominated
value F EM ! kc D (Fig. 2a). In the ‘‘maximal’’ end- by the sulfate reduction pathway, where the end products
member, the ecosystem effectively operates at thermody- of mineralization (H2 S, NH4 ) are not re-oxidized within
namic equilibrium, i.e. C R ¼ C W =K eq and d ! 0. In the sediment (Reimers et al., 1990; Meysman et al., 2003).
summary, when the resource processing parameter kEM In contrast, the bottom water in Aarhus Bay and Young
varies over the semi-infinite range ½0; 1, we find that the Sound is well oxygenated. As a result, the sediment is
ecosystem metabolism F EM varies over a finite range inhabited by a diverse macrofaunal community, and the
½0; kc D, and the corresponding disequilibrium distance d associated organic processing within these two sites is
varies over the finite range ½0; D. governed by an interplay between macrofaunal sediment
Table 1 shows that all three sites are characterized by reworking and microbial metabolism. Although sulfate
fast food webs. For Aarhus Bay and Young Sound, the reduction is still the dominant pathway, the reduced by-
observed resource processing parameter kdata is an order of products (H2 S, NH4 ) are now re-oxidized with O2 , and
magnitude higher than the threshold value kc . In these two thus the food web extracts more free energy out of the
sites, the actual ecosystem metabolism F data
EM is only slightly available organic matter.
lower than the limiting value F max
EM ¼ kc D. To show this, we
calculated the resource conversion efficiency, defined 4.3. Total and ecosystem entropy production
as Z ¼ F data max data
EM =F EM where F EM is derived from kdata via
Eq. (39). We find values of respectively 94% and 93% for By scanning the full range 0okEM o1, we have
AB and YS (Table 1). Note that these resource conversion examined all resource processing regimes that are possible
efficiencies are remarkably similar despite the very different for each of the three sites. We can now examine whether
setting (a temperate coastal bay versus an arctic sound). among these many possible regimes, there are any ‘‘remark-
The resource processing in the Santa Barbara Basin is able’’ ones from a thermodynamic perspective. To this end,
markedly ‘‘slower’’, as the observed resource processing we can use Eqs. (27) and (28) to calculate the entropy
parameter kdata is only twice as high as the threshold value production rates sEM and stot associated with a certain
kc , while the associated resource conversion efficiency ERPR. These relations are plotted in Fig. 3 for the three
attains 72%. The strong difference between the Santa sites. The characteristic shape of the two entropy production
Barbara Basin on the one hand, and the Aarhus Bay and rates are very different. The stot curves monotonically
Young Sound sites on the other hand, presumably reflects increase with increasing kEM . This contrasts strongly with
the intrinsic differences in food web structure and organic the unimodal sEM curves, which start at zero, go through a
matter processing at these sites. The Santa Barbara Basin is maximum, and subsequently decrease again to zero.
an anoxic basin, where most of the time, the bottom water The monotonic increase of stot is readily explained by
experiences suboxic to anoxic conditions (less than expression (28). The logarithmic factor remains constant,
0.01 mM O2 , Reimers et al., 1990). Due to this oxygen and so, the total entropy production rate simply scales with
deficiency, no macrofauna are present within the sediment the ecosystem metabolism F EM . The origin of the extremum
and so the benthic food web is entirely microbial (bacteria, in the sEM curve is explained by ‘‘minimal’’ and ‘‘maximal’’
ARTICLE IN PRESS
end-member situations discussed above. In both end-

Aarhus Bay
35 members the entropy production sEM vanishes because the
flow F EM acts oppositely to the thermodynamic force X EM .
Entropy production (μW m-2 K-1)
30
Total EP Metabolic EP In the ‘‘minimal’’ end-member (slow food web situation),
25
the concentration difference between the internal ecosystem
reservoirs is greatest (large but finite force), but the
20 ecosystem’s metabolism vanishes (zero flow). In contrast,
in the ‘‘maximal’’ end-member, the ecosystem metabolism is
15 maximal (large but finite flow), but concentration difference
between the internal reservoirs vanishes (zero force). In both
10 situations, the product of force and flow makes that sEM
vanishes. In between the end-member regimes, the entropy
5
κdata κmep production associated with ecosystem metabolism is posi-
0 tive, and hence, it goes through a maximum.
10-4 10-2 100 102 This state of ‘‘maximum ecosystem entropy production’’
Resource processing rate κEM (yr-1) forms a characteristic ERPR in the possible range
0okEM o1. To calculate the associated kmep , we can first
rewrite sEM from expressions (35) and (36), to obtain
12 Young Sound
kEM D
sEM ¼
1 þ kEM =kc
10 Total EP Metabolic EP !
K eq aW aR C 0R ð1 þ kEM =kc Þ kEM D
R ln .
8 aR aW C 0W ð1 þ kEM =kc Þ þ kEM D
ð45Þ
6
This expression has a single maximum, and hence, it shows
4 that the ecosystem entropy production sEM is unimodal in
the parameter kEM . The resource processing parameter
2 kmep at maximal entropy production (MEP) is obtained by
κdata κmep
solving the non-linear equation dsEM =dkEM ¼ 0. The
0
resulting values are listed in Table 1 for the three sites.
10-4 10-2 100 102 These values of kmep are compared to the reported value
Resource processing rate κEM (yr-1) kdata in Fig. 3a–c. Quite intriguingly, the actual values of
the resource processing parameter closely match the MEP
4 values for the Aarhus Bay and Young Sound sites. Again,
Santa Barbara Basin
the Santa Barbara Basin shows a distinct response, where
3.5
the actual resource processing kdata is considerably smaller
Total EP Metabolic EP than the theoretical MEP value. Accordingly, one can ask
3
why this difference between SBB and the two other sites?
2.5 One assumption in our calculations is that the same
equilibrium constant K eq and the same thickness L applies
2 to all sites. There is however no a priori reason why these
1.5
two parameters should be constant across sites. In contrast,
from a biogeochemical perspective, one would expect the
1 Santa Barbara Basin to be different. The value K eq ¼ 108
was based upon the aerobic respiration reaction (24). But
0.5 as discussed above, the free energy extracted for organic
κdata κmep
matter decomposition should be lower in the Santa
0
10-4 10-2 100 102 Barbara Basin, due to the absence of O2 -mediated re-
Resource processing rate κEM (yr-1)
oxidation reactions. Biogeochemically, one thus would
expect a lower K eq value in the SBB (sulphate reduction)
Fig. 3. The total entropy production stot and the metabolic entropy when compared to the other two sites (aerobic respiration).
production sEM as a function of the specific metabolic rate kEM for three Moreover, the value of L ¼ 10 cm was based on the global
marine sediment sites (Aarhus Bay, Young Sound, Santa Barbara Basin). average thickness of the bioturbated zone. However, the
The kdata value indicates the observed decay constant of organic matter,
kdata indicates the value predicted by entropy production maximization.
SBB does not experience bioturbation, and intense micro-
bial activity seems still present at greater depth into the
sediment (Meysman et al., 2003). To test these aspects, we
ARTICLE IN PRESS
1.5
1
0.8
1
κmep (yr-1)
κmep (yr-1)
0.6
κdata
0.4 0.5 κdata
0.2
0 0
5 10 15 20 100 105 1010 1015 1020
Sediment thickness L (cm) Equilibrium constant Keq
Fig. 4. Uncertainty analysis. The baseline simulation uses the Aarhus Bay setting with L ¼ 10 cm and K eq ¼ 108 . (a) MEP resource processing
parameter as a function of the sediment thickness. (b) MEP resource processing parameter as a function of the equilibrium constant. The dashed line
indicates the observed decay constant kdata .
conducted an additional simulation for the Santa Barbara floor. This approach makes abstraction of all internal
Basin with modified parameters K eq ¼ 102 and L ¼ 20 cm. details of the food web, and principally focuses on the
This provided a new value of kmep ¼ 0:073, which is now interaction of the food web with the external surroun-
only slightly larger than the observed value kdata ¼ 0:056. dings. Our analysis shows that in the steady-state, an
The parameter adjustment of the Santa Barbara Basin arbitrarily complex food web can be treated as a black
remains to some extent arbitrary: although the direction of box, represented by the parameter kEM , which condenses
the parameter adjustment can be biogeochemically justified all information on resource processing by the food
(lowering K eq and increasing L), the magnitude of the web. This analysis in terms of kEM provides a generic
adjustment remains arbitrary. As noted above, there are no way to assess the thermodynamic functioning of food
strict data constraints on the parameters L and K eq , and as webs, resulting in a gradient from slow food webs
a consequence, there is substantial uncertainty on the ðkEM ! 0Þ to fast food webs ðkEM ! 1Þ. If one specifies
associated values that were used in the above simulations. a specific food web structure, one effectively selects a
To investigate how sensitive kmep is to these two certain point along the axis 0okEM o1. In a case
parameters, we performed an uncertainty analysis, where study of three marine sediments, we found that actual
L was varied over the depth range ½5; 20 cm and the resource processing all fell with in the fast food web
equilibrium constant was varied over 20 orders of category, with oxic sediments showing faster resource
magnitude. Fig. 4 provides the resulting plots for the processing than anoxic environments. Further research
Aarhus Bay site settings. The MEP resource processing needs to determine whether this trend is as general as it
kmep decreases with sediment thickness, and increases with appears from this three site sample. If confirmed, natural
higher equilibrium constant. For both ranges, the variation oceanic sediments emerge as transport-limited systems,
in the kmep is about 1 order of magnitude. Accordingly, the where natural selection would favor various ways of
uncertainty on the kmep values in Fig. 3 is about a factor of enhancing the exchange of resource and waste products
10 higher or lower than the values depicted. This with the environment (i.e. biological feedbacks on aR
uncertainty is substantial, but is far less than the natural and aW ). Such transport enhancing activities are actually
variability kdata . In oceanic sediments, kdata ranges over omnipresent within the marine benthos, in the form of
more than eight orders of magnitude from 106 to 102 yr1 filter-feeding and bio-irrigation (Meysman et al., 2006).
between coastal and deep sea sediments (Boudreau, 1997; Our analysis thus could provide a thermodynamic per-
Middelburg and Meysman, 2007). Accordingly, we esti- spective on the evolutionary development of such transport
mate the uncertainty connected to kmep at about 15% of the enhancing strategies.
possible natural range of kEM . The fact that the observed
values kdata fall within the uncertainty range of kmep for all 5.2. Conceptual problems in thermodynamic ecology
three sites is therefore an intriguing observation.
The idea that thermodynamic constraints could play a
5. Discussion and conclusion role in ecosystem development, has recently instigated
considerable research into the use of thermodynamic
5.1. A thermodynamic perspective on the ocean floor principles in ecosystem analysis (Jorgensen, 2000; Jorgen-
ecosystem sen and Svirezhev, 2004). Yet, some concerns have been
issued about the quantitative rigor, the consistency, and the
We have developed a thermodynamic approach for heuristic nature with which thermodynamic concepts are
the chemotrophic, detritus-based ecosystem of the ocean employed within thermodynamic ecology (Mansson and
ARTICLE IN PRESS
McGlade, 1993; Wilhelm and Bruggemann, 2000; De Wit, and (b) the total entropy production stot within both
2005; Gaucherel, 2006). Therefore, in our analysis, we ecosystem and environment. Our analysis (Fig. 3) shows
devoted a lot of attention—almost fanatically—to a that only the ecosystem entropy production sEM has the
rigorous mathematical definition of thermodynamic con- proper mathematical form of a consistent goal function. The
cepts (e.g. thermodynamic gradient, disequilibrium dis- total entropy production stot strictly increases with the
tance, entropy production, physical environment, ERPR) resource conversion rate kEM , and hence, it does not show a
and to their correct translation into ecological terms. Every maximum. In the real world, maximizing stot would select a
concept that was employed, was also defined in a strict state of maximal resource processing that would have a finite
mathematical sense. value for kEM . However, any other monotonic function in
Our analysis provides two clear examples of where past kEM would select the same state, and therefore, stot cannot
treatments have been confusing. A first issue is the often be regarded a proper discriminatory goal function.
quoted statement that ‘‘ecosystems are far-from-thermo- In contrast, the metabolic entropy production sEM does
dynamic-equilibrium’’. The classical interpretation of the show a clear maximum when the resource processing rate
distance from equilibrium is based on the difference in the varies from zero to infinity. Because sEM has the right
chemical potential between biomass and some reference mathematical form, this does not mean that the ecosystem
state. We show that this interpretation is troublesome, entropy production also acts as a true ecological goal
because (1) no procedure exists yet to define the chemical function. Whether sEM means something for ecosystem
potential of biomass, and (2) no proper ‘‘natural’’ reference development, has to be verified from experimental ob-
state can be defined. To resolve this situation, we propose servations on ecosystem functioning. To this end, we
two alternative measures for the distance from equilibrium quantified entropy production in three marine sediment
(D and d), which are based upon the disequilibrium ecosystems. More specifically, we compared values for the
between resources and waste products in respectively resource processing metric kdata as reported from biogeo-
environment and ecosystem. chemical studies with the value predicted by the maximiza-
A second problem concerns the use of the term ‘‘entropy tion of the ecosystem entropy production sEM . The kmep
production’’ in the context of ecological goal functions. values approach the kdata within one order of magnitude,
Such goal functions are extremal principles, which have which is intriguing, given that kdata in sediments can
been proposed as relevant indicators for ecosystem vary over eight orders of magnitude (Middelburg and
development—see Muller and Leupelt (1998) for a critical Meysman, 2007). However, based on our limited sampling
review of the predictive capabilities. In connection to such of three sediments, and given the uncertainty on kmep due
ecological goal functions, the notion of ‘‘maximal entropy to uncertain estimates for L and K eq , this study does not
production’’ has surfaced a number of times in recent years provide unequivocal evidence that ecosystem entropy
(Ulanowicz and Hannon, 1987; Schneider and Kay, 1994; production sEM does function as a goal function. The
Svirezhev, 2000; Fath et al., 2001; Schneider and Sagan, correspondence between kdata and kmep could be a mere
2005). Unfortunately, the term ‘‘entropy production’’ has coincidence. Still, our results are intriguing enough to
always been used without a proper definition of what this stimulate further investigation into the link between
rate actually encompasses. In other words, it is not entropy production and resource processing in food webs.
specified which interactions are included in the entropy Such investigations should focus on more marine sediment
production rate. Our analysis shows that such an sites to see whether the pattern holds, and preferable, very
unconstrained use of the term ‘‘entropy production’’ is different ecosystems as well (e.g. terrestrial soils).
problematic. Multiple entropy production rates can be
defined, depending on what interactions are accounted for 5.4. Thermodynamic analysis of other ecosystems
in the entropy production budget (sR , sW , sA , sM , and sT
are the possible contributions in steady-state for a single- Our analysis shows that a proper investigation of
consumer ecosystem). These various entropy production ‘‘gradients’’ is needed to calculate the entropy production
rates may show different extremal behavior. within ecosystems. Such gradients exist within the ecosys-
tem ðdÞ, and between system and environment ðDÞ. Once a
5.3. Entropy production as an ecological goal function? consistent calculation procedure for the entropy produc-
tion is established, one can examine whether or not the
Given previous propositions along this line (Ulanowicz entropy production is a relevant measure for the behavior
and Hannon, 1987; Schneider and Kay, 1994; Svirezhev, of food webs. Accordingly, a pertinent question is whether
2000; Schneider and Sagan, 2005), it is logical to question and how our thermodynamic model analysis of detrital-
whether entropy production serves as a useful ecological goal based ecosystems can be extended to other ecosystem
function? And if so, which entropy production rate should be types.
maximal? Among the many possibilities, we have identified In our description, the ecosystem operation depends on
two entropy production rates that could be relevant for three basic interactions: (1) the supply of resources from a
ecosystem functioning: (a) the entropy production sEM high-quality reservoir in the environment to the ecosystem,
associated with resource conversion within the ecosystem, (2) the conversion of resources into waste products within
ARTICLE IN PRESS
the ecosystem, and (3) the discharge of waste products into References
a low-quality reservoir in the environment. Each of the
three steps is ‘‘dissipative’’ in the sense that each step leads Aoki, I., 1995. Entropy production in living systems—from organisms to
to a corresponding positive entropy production. The true ecosystems. Thermochim. Acta 250, 359–370.
Aoki, I., 2001. Entropy and exergy principles in living systems. In:
dissipative effect of the food web is the entropy production Jorgensen, S.E. (Ed.), Thermodynamics and Ecological Modelling.
sEM associated with step (2), i.e., the resource conversion Lewis Publishers, Boca Raton, pp. 167–190.
that allows the food web to maintain its self-organization. Aoki, I., 2006. Ecological pyramid of dissipation function and entropy
Our analysis of ocean floor sediments indicates that from production in aquatic ecosystems. Ecol. Complexity 3, 104–108.
all entropy production rates, sEM surfaces as the only Battley, E.H., 1999. On entropy and absorbed thermal energy in biomass;
a biologist’s perspective. Thermochim. Acta 331, 1–12.
candidate that could have ecological significance.
Battley, E.H., Putnam, R.L., Boeriogoates, J., 1997. Heat capacity
If one is able to properly map resources and waste, the measurements from 10 to 300 K and derived thermodynamic functions
above three-step-mechanism could apply to chemo-auto- of lyophilized cells of saccharomyces cerevisiae including the absolute
trophic and photo-autotrophic ecosystems as well, and so it entropy and the entropy of formation at 298.15 K. Thermochim Acta
could provide a road map to eventually calculate sEM . In 298, 37–46.
the case of chemo-autotrophy, one can think of a marine Begon, M., Harper, J.L., Townsend, C.R., 1996. Ecology. Blackwell
Science, Oxford.
hydrothermal vent system, where (1) reduced chemical Berg, P., Rysgaard, S., Thamdrup, B., 2003. Dynamic modeling of early
compounds (e.g. H2 S) percolate out of the sea floor (the diagenesis and nutrient cycling. A case study in an arctic marine
high-quality reservoir in the environment), (2) these sediment. Am. J. Sci. 303, 905–955.
reduced species are then oxidized ðH2 S ! SO4 Þ, and the Boudreau, B.P., 1997. Diagenetic Models and Their Implementation.
free energy of this redox reaction is used to build and Springer, Berlin.
Boudreau, B.P., 1998. Mean mixed depth of sediments: the wherefore and
maintain the food web of the vent system, and (3) the
the why. Limnol. Oceanogr. 43, 524–526.
oxidized products ðSO4 Þ are then mixed within the over- Burdige, D.J., 2006. Geochemistry of Marine Sediments. Princeton
lying water column. The gradients D and d that character- University Press, Princeton, NJ.
ize the hydrothermal vent ecosystem are based on free Canfield, D.E., Thamdrup, B., Kristensen, E., 2005. Aquatic Microbial
energy difference between reduced and oxidized species. Ecology. Elsevier, San Diego.
Just like the heterotrophic ecosystems studied here, chemo- Case, T.J., 2000. An Illustrated Guide to Theoretical Ecology. Oxford
University Press, New York.
autotrophic ecosystems fit within the framework of de Ruiter, P., Wolters, V., Moore, J.C., 2005. Dynamic Food Webs:
standard chemical thermodynamics, and so we expect that Multispecies Assemblages, Ecosystem Development, and Environ-
descriptions will be rather similar. mental Change. Elsevier, Amsterdam.
However, things appear more complicated in photo- De Wit, R., 2005. Do all ecosystems maximise their distance with respect
autotrophic ecosystems. This is because one needs to to thermodynamic equilibrium? A comment on the ‘‘ecological law of
thermodynamics’’ (Elt) Proposed by Sven Erik Jorgensen. Sci. Mar.
invoke the more complex thermodynamics of radiation, in
69, 427–434.
addition to chemical thermodynamics. Still, the general Fath, B.D., Patten, B.C., Choi, J.S., 2001. Complementarity of ecological
three-step scheme also holds for ecosystems that are driven goal functions. J. Theor. Biol. 208, 493–506.
by photosynthesis. In this case, (1) radiation energy is Fossing, H., Berg, P., Thamdrup, B., Rysgaard, S., Munk Sorensen, H.,
derived from a high quality reservoir (the sun at 5800 K), Nielsen, K., 2004. A model set-up for an oxygen and nutrient flux
(2) short-wavelength solar radiation is converted to long- model for Aarhus bay (Denmark). NERI Technical Report No. 483.
Gaucherel, C., 2006. Influence of spatial patterns on ecological applica-
wavelength heat radiation in photosynthesis, (3) the tions of extremal principles. Ecol. Model. 193, 531–542.
resulting heat is subsequently radiated back to a low- Glansdorff, P., Prigogine, I., 1971. Structure, Stability, and Fluctuations.
quality reservoir (deep space at 3 K). At first sight, the Wiley, New York.
thermodynamic gradient D in photo-autotrophic ecosystem Grover, J.P., 1997. Resource Competition. Chapman & Hall, London.
should involve the wavelength difference between the Jorgensen, S.E., 2000. Thermodynamics and Ecological Modelling. Lewis
Publishers, Boca Raton.
radiation emitted by the sun and deep space. However,
Jorgensen, S.E., Svirezhev, Y.M., 2004. Towards a Thermodynamic
the infilling of d is less obvious. Clearly, there is some Theory for Ecological Systems. Elsevier, Amsterdam.
analogy between chemotrophic and phototrophic systems, Kondepudi, D., Prigogine, I., 1998. Modern Thermodynamics: From Heat
but how far this analogy stretches must be resolved in Engines to Dissipative Structures. Wiley, Chichester.
future studies. Kristensen, E., Haese, R.R., Kostka, J.E., 2005. Interactions between
Macro- and Microorganisms in Sediments. American Geophysical
Union, Washington, DC.
Kurzynski, M., 2006. The Thermodynamic Machinery of Life. Springer,
Acknowledgements Berlin.
Lindeman, R.L., 1942. The trophic dynamic aspect of ecology. Ecology
We thank Rich Williams, Dick Van Oevelen, and two 23, 399–418.
anonymous reviewers for constructive input on an earlier Loreau, M., Holt, R.D., 2004. Spatial flows and the regulation of
ecosystems. Am. Nat. 163, 606–615.
manuscript version. This research was supported by a grant
Mansson, B.A., McGlade, J.M., 1993. Ecology, thermodynamics and
from Netherlands Organization for Scientific Research Odum, H.T. conjectures. Oecologia 93, 582–596.
(NWO PIONIER, 833.02.2002). This is publication 4101 of May, R.M., 1981. Theoretical Ecology. Blackwell Scientific Publications,
the Netherlands Institute of Ecology (NIOO-KNAW). Oxford.
ARTICLE IN PRESS
Meysman, F.J.R., Middelburg, J.J., Herman, P.M.J., Heip, C.H.R., 2003. Schneider, E.D., Kay, J.J., 1994. Life as a manifestation of the 2nd law of
Reactive transport in surface sediments. II. Media: an object-oriented thermodynamics. Math. Comput. Model. 19, 25–48.
problem-solving environment for early diagenesis. Comput. Geosci. Schneider, E.D., Sagan, D., 2005. Into the Cool: Energy Flow,
29, 301–318. Thermodynamics and Life. The University of Chicago Press,
Meysman, F.J.R., Middelburg, J.J., Heip, C.H.R., 2006. Bioturbation: a Chicago.
fresh look at Darwin’s last idea. Trends Ecol. Evol. 688–695. Schroedinger, E., 1944. What is Life? Cambridge University Press,
Middelburg, J.J., Meysman, F.J.R., 2007. Ocean science—burial at sea. London.
Science 316, 1294–1295. Solé, R.V., Bascompte, J., 2006. Self-Organization in Complex Ecosys-
Montoya, J.M., Pimm, S.L., Sole, R.V., 2006. Ecological networks and tems. Princeton University Press, Princeton.
their fragility. Nature 442, 259–264. Susani, L., Pulselli, F.M., Jorgensen, S.E., Bastianoni, S., 2006.
Morowitz, H.J., 1978. Foundations of Bioenergetics. Academic Press, Comparison between technological and ecological exergy. Ecol.
New York. Model. 193, 447–456.
Muller, F., Leupelt, M. (Eds.), 1998. Eco Targets. Goal Functions, and Svirezhev, Y.M., 2000. Thermodynamics and ecology. Ecol. Model. 132,
Orientors, Springer-Verlag, Berlin. 11–22.
Nicolis, G., Prigogine, I., 1977. Self-Organization in Nonequilibrium Tilman, D., 1982. Resource Competition and Community Structure.
Systems. Wiley, New York. Princeton University Press, Princeton.
Odum, H.T., 1983. System Ecology. Wiley Interscience, New York. Ulanowicz, R.E., Abarca-Arenas, L.G., 1997. An informational synthesis
Reimers, C.E., Lange, C.B., Tabak, M., Bernhard, J.M., 1990. Seasonal of ecosystem structure and function. Ecol. Model. 95, 1–10.
spillover and varve formation in the Santa-Barbara Basin, California. Ulanowicz, R.E., Hannon, B.M., 1987. Life and the production of
Limnol. Oceanogr. 35, 1577–1585. entropy. Proc. R. Soc. Lond. Ser. B—Biol. Sci. 232, 181–192.
Sandler, S.I., Orbey, H., 1991. On the thermodynamics of microbial- Wilhelm, T., Bruggemann, R., 2000. Goal functions for the development
growth processes. Biotechnol. Bioeng. 38, 697–718. of natural systems. Ecol. Model. 132, 231–246.

Termodinamica Ocean

Uploaded by

Copyright:

Available Formats

Termodinamica Ocean

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Termodinamica Ocean

Uploaded by

Copyright:

Available Formats

ARTICLE IN PRESS

Journal of Theoretical Biology 249 (2007) 124–139

A thermodynamic perspective on food webs: Quantifying entropy

Keywords: Non-equilibrium thermodynamics; Ecology; Modeling; Dissipation; Ecosystem

1. Introduction Thermodynamics offers a different, complementary per-

we obtain 2.5. Entropy balance formulation

Input data AB YS SBB

Burial velocity o cm yr1 0.32 0.12 0.26

Resource concentration (mol m-3)

Waste concentration (mol m-3)

slow food web fast food web

end-member situations discussed above. In both end-

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.