Istituto Nazionale di Oceanografia e di Geofisica Sperimentale OGS

1
SESAME International Summer Course on Coupled
Ecological Modelling 8-13 June 2009
Theory of Ecopath with Ecosim and applications in
the Mediterranean and Black Sea
Simone Libralato
Istituto Nazionale di Oceanografia e di Geofisica
Sperimentale (OGS) Dept. Oceanography - Trieste,
Italy e-mail slibralato_at_ogs.trieste.it
2
OGS (Trieste-Italy)
National Istitute of Oceanography and
Experimental Geophysics OGS
Dept. Oceanography
ECHO group - Ecology and Computational
Hydrodynamics in Oceanography
(MODELLING MARINE ECOSYSTEMS)
Who
What
Hydrodynamic models Trophodynamic
models Biogeochemical models Food web
models Individual-Based models Large Eddy
simulations Data Assimilation Sensitivity analysis
Cosimo SOLIDORO Head Donata MELAKU CANU
Scientist Gianpiero COSSARINI Scientist Simone
LIBRALATO Scientist Stefano SALON
Scientist Vinko BANDELJ res fellow Paolo
LAZZARI res fellow Stefano QUERIN res
fellow Anna TERUZZI PhD student Anna
SUSTERSIC PhD student Valentina MOSETTI
Technician Tomaso FORTIBUONI external
collaborator Sebastiano TREVISANI external
collaborator
3
ECHO group
Where and why
Water quality Nutrient enrichment
Eutrophication Primary production Water pollution
(oil spill) Bioaccumulation of toxicants Effects
of fishing Aquaculture impacts Climate change
Trieste
Mediterranean Sea
4
ECHO group Other activities
Session Modelling marine systems http//www.isem
na.org/
5
Theory of Ecopath with Ecosim and applications

• OUTLINE
• General introduction to Ecopath with Ecosim
• Ecopath theory input parameters balancing the
  model
• Outputs of Ecopath synthetic indicators
  trophic flows
• Ecosim (0D, time dynamic) additional
  parameters forcing and mediation functions
  fishing dynamics
• Brief introduction to Ecospace

6
Ecopath with Ecosim for an Ecosystem approach
Food web modelling (ECOPATH with ECOSIM EwE,
Atlantis)
Trophic level
Pauly et al., 1998
Biological detail/articulation
Direct and Indirect effects, cascading effects,
unpredicteable effects, trade-offs....
7
A general key issue before going on
A model is a representation of the reality .
Art is a lie that helps us to realize the
truth. (Pablo Picasso) ..the same for modelling!
reality
model
The model captures essential features/particular
aspects of reality. Thus for the same reality
there can be different models for different
purposes
8
Ecopath with Ecosim
www.ecopath.org

• Ecopath with Ecosim (EwE) is a free
  ecological/ecosystem modeling software suite. EwE
  has three main components Ecopath - a static,
  mass-balanced snapshot of the system Ecosim - a
  time dynamic simulation module for policy
  exploration and Ecospace - a spatial and
  temporal dynamic module primarily designed for
  exploring impact and placement of protected
  areas. The Ecopath software package can be used
  to
• Address ecological questions
• Evaluate ecosystem effects of fishing
• Explore management policy options
• Analyze impact and placement of marine protected
  areas
• Predict movement and accumulation of contaminants
  and tracers (Ecotracer)
• Model effect of environmental changes.

9
EwE applications in the Mediterranean Sea
Venice lagoon
Gulf of Trieste

Adriatic Sea

Black Sea
Corsica
Orbetello lagoon
Catalan Sea
Murcia coast
Aegean Sea
Ionian Sea

Explore the trophic structure and apply network
analysis at different scales Examine ecosystem
effects induced by fishing Explore the
consequence of species introduction Analyze
management questions MPA, selectivity
devices Explore the effects of environmental
drivers on ecosystem dynamics Explore
bioaccumulation of organic pollutants in the food
web Develop end-to-end models coupling high to
low TL models Develop ecosystem indicators
keystone species indicators and the loss in
production







10
EwE modelling the food web with functional groups
EwE food web is represented by means of
functional groups (autotrophs, consumers,
detritivores, etc) linked by biomass consumption
flows (predation). Represented also respiration,
detritus flows and exports as catches.
Ecosystem

• A functional group can represent
• a portion of the population (e.g. juveniles)
• a species
• group of species ecologically similar (same
  predators, same preys, similar consumption and
  predation rates)

11
How many functional groups?
Describe all trophic levels introduce at least
one detritus group (mandatory) describe with one
functional group interesting species
(commercially important or key species).
Phanerogames/algae
Plankton
Example the Lagoon of Venice the food web of
is described by means of 27 functional groups
Macrobenthos
Venice lagoon
Fishes
Birds
Adriatic Sea
Non-living groups
(Pranovi et al., 2003, Marine Biology)
12
1) ECOPATH (0D, NO time dynamic)
Mass or energy balance primary producers
Flow of energy And/or biomass
Fishing yield
Biomass accumulation
Predation
Production
Migration
Functional Group (i)
Other mortality
(Christensen et al., 2005, User Guide)
13
1) ECOPATH (0D, NO time dynamic)
Mass or energy balance consumers
Flow of energy And/or biomass
respiration
Fishing yield
Biomass accumulation
Predation
Consumption of prey
Production
Migration
Functional group (k)
Other mortality
Non-assimilated
14
A particular input parameter
The Ecotrophic Efficiency
The Ecotrophic efficiency is the fraction of
production of the functional group that is used
as predation and fishing yield within the
DESCRIBED food web. It is a parameter that is
highly dependent from the description of the food
web better to leave it as unknown (to be
estimated by EwE).
Food web 1
Seabass is the top predator, there is no fishing
on it nor predation by seabirds
EE0
P0.57
Production and consumption rates as year-1
15
Mass-balance system of equations

• Quasi Steady-state, in order to have
• - Biomasses almost constant
• average annual rates
• Biomass accumulation close to zero ?B ? 0

• ADVANTAGES
• lower number of param
• average rates are constant
• DISADVANTAGES
• - Loose of some realism
• no seasonality
• USEFULNESS
• base of a dynamic model
• medium term analysis

Period 2
Biomass
Period 3
Period 1
Time
16
Mass-balance system of equations
Primary producers
Consumers
The system of algebric equations link the
functional groups by means of the predation terms
(Q/B)j Bj DCij solving the system allow for
estimating at least one parameter for each
functional group (parsimony of parameters).
In Ecopath the user can decide to have one of the
parameters in cian to be estimated by the model
which parameter (choice driven by data
available) some parameters (in red) have to be
necessarily introduced fishing yield and diet.
Moreover in Eq. 2 respiration and Not assimilated
17
Input parameters
s
h
Unit for biomass wet weight, dry weight, carbon,
energy (Joule, calorie) Rates monthly or yearly
basis
(Christensen et al., 2005, User Guide)
18
Introducing input parameters
1)Basic input biomasses, production and
consuption rates. Empty cellsparameters to be
estimated
2) Diet composition
4)Fishery one can introduce landings,
discards, prices, costs, flow of discard ect
3) Detritus fate here to describe where deads
and excretions go.
19
then is the model BALANCED?
Data introduced might be inconsistent!
? - two fundamental constrains need to be met

• all EEi1
  
  (Energy cannot be created or distroyed, just
  transformed)
• Ri0
• (All energy transformations occour with
  losses)
• A consequence of 2) is that the consumption
  efficiency gi0.5 (but for some groups, such as
  bacteria)

• If the model is not balanced it is necessary to
  revise the input parameters
• look at errors in the diet matrix
• pay attention to species that feed outside the
  system or move into the system for some periods
  (es. migrants)
• revise production and consumption rates
• maybe the system is not at steady-state BA0?

20
Example of an ECOPATH food web model the Lagoon
of Venice
Piscivorous birds
gobies
Seabass
Sand smelt
Seabream juv
Seabream
Seabass juv
Mullets juv
Macrob. carnivorous
Benthic predators
Mullets
Macrob. omnivorous
Zooplankton
Filter feeders
Macrob detritivorous
Clam juv
Bacterioplankton
Manila clam
Macrob herbivorous
Meiobenthos
Bottom sediment
SOM
Phytoplankton
Seagrass
Ulva
Other macroalgae
Epiphytes
(Libralato et al., Mar Ecol PSZNI, 2002 Pranovi
et al., Mar Biol, 2003)
21
Outputs of a balanced Ecopath model
Indicators estimated for functional groups
When the food web is balanced all the flows
related to each functional group are determined
(and coherent with those of the other groups)
Migration
Onthe basis of these flows Ecopath can estimate
indicators (features/characteristics) for each
functional group, such as
Respiration
Yield
Biomass accumulation
Prey consumption
Predation
Production
Not assimilated
Flow to detritus

• - Trophic level (TLi)
• Omnivory index (OIi)
• Trophic impacts (MTIij)
• niche overlap
• Keystone species (KSi)
• Primary Production Required (PPRi)
• Ascendency

22
Example of output estimating direct/indirect
effects
A
B
dBA
dCB
fBA
fCB
According to Ulanowicz and Puccia (1990) dij
fraction of the prey i in the diet of the
predator j fij fraction of total consumption of
i used by predator j the net impact of i on j
(qij) is given by the difference between positive
effects (d) and negative effects (f) qij
dij-fji
23
Example of output estimating direct/indirect
effects
The total impact along a series of nodes i,j,k
,y,z is qij qjk qyz Therefore the matrix
of qij , i.e. Q, one obtains the matrix M
called mixed trophic impact whose elements are
mij
impacted
impcting
24
Example of output estimating the keystone groups
1 Birds 2 Macrobenthos mixed-feeders 3
Macrobenthos herbivorous 4 Zooplankton 5
Micro-Meiobenthos 6 Sand smelt 7 Macrobenthos
detritivorous 8 Manila clam comm. 9 Grass goby 10
Phytoplankton 11 Macrobenthos carnivorous 12
Macrobenthos filter-feeders 13 Epiphytes 14
Bacterioplankton 15 Seabass ad. 16 Manila clam
juv. 17 Nekton carnivorous benthic feeders 18
Other Macroalgae 19 Ulvales 20 Mullets ad. 21
Seabream juv. 22 Seabass juv. 23 Seagrasses 24
Seabream ad.
25
Synthetic indicators of the food web
Ecopath allows to estimate total flows of the
ecosystem
Total system troughoutput (T) Sum of respiration
(Rtot) Sum of consumption (QTOT) Sum of
producitons (Ptot) Total flows to detritus
(Fdet) Total Primary Production (PPtot) Total
biomass of the web (Btot) Total catches (Ctot)
And then one can estimate synthetic indicators
for the food web

• Ptot/Rtot
• PPtot/Btot
• Ctot/PPtot
• Btot/T
• total flows of recycling
• cycling index (Finn index, FCI)
• mean number of energy paths

• System Omnivory Index (SOI)
• Connettivity Index (CI)
• Ascendency
• Transfer Efficiency (TE)
• Primary prodution required (PPR)
• mean TL of catches
• .

(Odum, Science, 1969)
26
Comparing ecosystems
Protected area Riserva Naturale Marina di
Miramare Protected since 1986 Food web model 23
functional groups years represented
2000-2003 Coastal area
Exploited area North and Central Adriatic
Sea heavily exploited (bottom trawl, beam trawl,
mid-water trawl, purse seine, tuna
fisheries) Food web model 40 functional
groups years represented 1990s Excludes 3 NM
from the coast
(Libralato et al., ICES CM, 2005 Coll et al.,
2007)
27
Synthetic indices protected vs fished
Synthetic indices
Exploited
Protected
(Libralato et al., ICES CM, 2005)
28
Ecopath models base for meta-analyses
Testing new indices
o Group 1- sustainably fished ecosystems (23
models) Group 2- overexploited ecosystems (32
models)
(Libralato et al., MEPS, 2008 Coll et al.,
PlosOne, 2008 Mora et al., 2009, PlosBiology)
29
2) Ecosim the dynamic routine of EwE
From Ecopath (0D, no time dynamic) to Ecosim
Ecopath
From the system of algebric equation ot the
system of differential equations
Ecosim
Immigration
Production term
Mortality (natural, fishing)
Predation
Walters et al. 1997. Rev. Fish Biol. and Fish.,
7 139-172
30
Ecosim with respect to Ecopath, another class of
model
The structure of the food web and many parameters
of ECOPATH are used to derive parameters of the
dynamic model, ECOSIM.

• What ECOSIM takes from ECOPATH
• set up a system of differential equations from
  the baseline Ecopath model
• use initial conditions (biomasses) from Ecopath
• growth efficiency rate (gi Pi/Qi) and initial
  mortality rates (F,M0) from Ecopath

Functional responses make ECOSIM different from
other dynamic models.
Walters et al. 1997. Rev. Fish Biol. and Fish.,
7 139-172
31
Functional responses
Functional responses
Cij prey killed
Lotka-Volterra
Bi prey density
Bj predator density
Cij prey killed
Holling type II
Bj predator density
Bi prey density
32
Ecosim the foraging arena
Functional response in ECOSIM the foraging
arena theory
Predator-prey encounter is NOT random in space
due to behavioural and physical mechanisms
(hiding, schooling, etc). Foraging arena only a
fraction of the prey biomass is available for
predator to consumption vulnerable biomass (Vij)
Biomass of the prey vulnerable
Vulnerability parameters
Walters et al. 1997. Rev. Fish Biol. and Fish.,
7 139-172 Walters et al. 2000. Ecosystems, 3
70-83 Christensen and Walters 2004. Ecol.
Model., 172(2-4) 109-139 Walters and
Christensen, 2007, Ecological Modelling
33
Ecosim the foraging arena
The resulting function is very flexible
This is the maximum mortality rate
Foraging arena
Cij prey killed
Bj predator density
Bi prey density
Walters et al. 1997. Rev. Fish Biol. and Fish.,
7 139-172 Christensen et al. 2000. EwE user
guide Christensen and Walters. 2004. Ecol.
Model., 172(2-4) 109-139
34
Ecosim parametrization
Ecopath with Ecosim user insert the scaling
factor (xij) between actual (Ecopath) mortality
rate and maximum
The values xij give an idea of how high is the
maximum mortality with respect to mortality in
Ecopath (default is 2) then having this xij and
initial values of biomasses and Qij, aij and vij
are calculated.
Ecopath mortality rate
Low values of xij (around 1)
Predator abundance has low effects on Consumption
(donor-driven)
Cij prey killed
High values of xij (gtgt20)
Lotka-Volterra behavior
Bj predator density
Walters pers. Comm Walters et al. 1997. Rev.
Fish Biol. and Fish., 7 139-172 Aydin et al.
2006. Deep Sea Research II, 52 757-780
35
Ecosim parametrization
Ecopath with Ecosim user insert the scaling
factor (xij) between actual (Ecopath) mortality
rate and maximum
The values xij give an idea of how high is the
maximum mortality with respect to mortality in
Ecopath (default is 2) then having this xij and
initial values of biomasses and Qij, aij and vij
are calculated.
Ecopath mortality rate
Low values of xij (around 1)
High values of xij (gtgt20)
Cij prey killed
Ecopath base biomass
Lotka-Volterra behavior
Relative change on predator density Bj
Walters pers. Comm Walters et al. 1997. Rev.
Fish Biol. and Fish., 7 139-172 Aydin et al.
2006. Deep Sea Research II, 52 757-780
36
Ecosim. Even more than that other effects
Time spent feeding and handling time
Time spent feeding is affecting predation
mortality. More time spent feeding more risk of
being predated. It is difficult to distinguish
between getting food and being food (C.J.
Walters).
Ti time spent feeding (for prey) Tj time
spent feeding (for predator) Dj effect of
handling time as a limit to consumption
Walters et al. 1997. Rev. Fish Biol. and Fish.,
7 139-172 Walters et al. 2000. Ecosystems, 3
70-83 Walters Kitchell, 2001. Can. J. Fish.
Aquat. Sci., 58 39-50
37
Ecosim. Even more than that other effects
Mediation function
Predator-prey interaction is affected by a third
species
j
i
K
Can be facilitation or protection of predation
Christensen et al. 2000. EwE user guide
Christensen and Walters. 2004. Ecol. Model.,
172(2-4) 109-139
38
Ecosim. Even more than that other effects
Forcing functions

• Physical or other environmental factors can
  influence species interaction.
• There are two main forcing functions applicable
  in EwE
• - seasonal may be applied to biomass production
  or egg production
• long term for represent decadal regime shifts
  or changes

Christensen et al. 2000. EwE user guide
Christensen and Walters. 2004. Ecol. Model.,
172(2-4) 109-139
39
Ecosim a complex parametrization
40
Dynamic simulations with Ecosim
ECOSIM A system of differential equations is
solved through time
Primary producers
...N...
Consumers
For the representation of dynamics of fisheries,
fishing rate (relative to initial Ecopath
estimates) can be varied during time for fleet or
for species
41
Set up of EwE model for the Catalan Sea
40 functional groups 4 Fishing
fleets Trawling purse seine long line troll bait
fleet
Coll, Palomera, Tudela, Dowd (2008) Ecological
Modelling 217 95116
42
Model hindcast (long time series)
Time series of biomass from 1978 to 2003 were
used for hindcast the model
Coll, Palomera, Tudela, Dowd (2008) Ecological
Modelling 217 95116
43
Model hindcast (long time series)
Time series of catches from 1978 to 2003
Coll, Palomera, Tudela, Dowd (2008) Ecological
Modelling 217 95116
44
3) ECOSPACE
ECOSPACE is a time dynamic spatially explicit
model, based on Ecosim.

• - 2D spatial domain in grid cells in each cell
  all Ecosim components
• In the 2D domain can be defined different
  habitats
• advecton field can be included
• for each species definition of preferred habitat
• migration patterns can be specified
• definition of Marine Protected Areas

Christensen and Walters 2004. Ecol. Model.,
172(2-4) 109-139
45
3) ECOSPACE
Assumptions

• Simmetrical movement from cell to adjectent
  ones, modified according to habitat preference
• Increase of predation risk in non preferred
  habitats
• Fishing effort is dynamically distributed
  according to fishing costs and revenues

Advantages and utilities
Predict spatial patterns of fisheries and
species Explore effects of management alternative
options (fishing effort allocation, MPA) Explore
spill over from MPA
46
Ecospace application for the Venice Lagoon
Ecopath with Ecosim food web
Biogeochemical 3D-NPZD calibrated model
The yearly average field of Dissolved Inorganic
Nitrogen (DIN) for a representative year (2001)
was obtained from TDM
TDM
Yearly average DIN field from TDM (year 2001)
used as input forcing parameter in the
spatio-dynamic routine of the Ecopath model
(ECOSPACE) 2 major habitats were defined and
food web components apply to opportune cells
Libralato et al., submitted
47
Results from 2D one-way coupling
Spatial distribution of functional groups
48
SESAME Integration of biogeochemical processes
in EwE
The integration of BGC results properly into EwE
is not a trivial task!!!

• EwE extended for including BGC processes
• include NUTRIENT as a functional group
• phytoplankton as consumer of nutrient
• impose effect of Temperature on plankton
  community as in the BGC model
• nutrient input obtained from the BGC model as a
  result of input-outputs at the boundaries

• HIGH TROPHIC LEVELS deads and excretions flows to
  organic matter, nutrient and sediment pools
• annual averages of degradation, sink and export
  of nutrient and detritus pools from BGC model

(Libralato et al., Eco Mod, submitted)
49
The modelling approach and the difficulties

• Temperature forcings
• This setting is done in accordance with
  formulations in BGC models (basin-average through
  time)

Forcing function
Time (years)
2) INPUT/OUTPUT of nutrient This setting is done
by using the basin-average nutrient input/output
resulting from the BGC model
Forcing function
EwE
3) Set correct functional relathionships as in
the BGC model for the overlapping functional
groups (zoo, phyto)
Ecopath
Consumption
BGC
Prey
50
Testing the Extended-EwE
(Cossarini Solidoro, Ecol Mod, 2008)
BGC
EwE
(Coll et al., J Mar Sys, 2007)
51
SESAME International Summer Course on Coupled
Ecological Modelling 8-13 June 2009
Thanks
Simone Libralato
Istituto Nazionale di Oceanografia e di Geofisica
Sperimentale (OGS) Dept. Oceanography - Trieste,
Italy e-mail slibralato_at_ogs.trieste.it