Online Resource 3. Main features of existing ecosystem modeling approaches
Table OR3.1 details the following features of existing ecosystem modeling approaches: (1) whether the modeling approach is spatial or non-spatial; (2) whether the modeling approach is steady-state or dynamic; (3) whether the modeling approach is trophodynamic, age-/size-/stage-structured or individual-based; (4) the time step of the modeling approach; (5) the ecosystem components represented in the modeling platform; (6) the fish processes represented in the modeling approach among: growth, survival, reproduction and movement; and (7) the calibration process of the modeling platform. Table OR3.2 indicates the other important features of existing ecosystem modeling approaches, as well as reference publications for each modeling approach.
1
Table OR3.1. Main features of existing ecosystem modeling approaches. N/A = not applicable
Ecosystem modeling approach / Spatial or non-spatial? / Dynamic or steady-state? / Tropho-dynamic, age-/size-/stage-structured or individual-based? / Time step / Ecosystem components / Fish processes represented among: growth, survival, reproduction and movement / Calibration process /Conceptual model / N/A / N/A / N/A / N/A / All the biological (fish and shellfish, marine birds, etc.) and human (activities) components of the ecosystem of interest / N/A / N/A
Loop analysis / Non-spatial / Steady-state / N/A / None / Diverse species and functional groups / None / No calibration
Extension of single-species assessment model (ESAM) / Non-spatial / Dynamic / Age-structured / 1 year / (1) One focal species;
(2) One or several predator species treated as "fisheries" (each having a "fishing effort", a catchability, and a selectivity-at-age function); or one or several environmental stressors treated as "fisheries" (each having a "fishing effort", a catchability, and a selectivity-at-age function). / Survival / Calibration using time series of fisheries catch, survey and fishery age composition estimates, and survey biomass estimates; and (1) predation-per-unit-predator-effort (p.p.u.e.) estimates; or (2) estimates of “catch-per-unit-stressor-effort”.
Extension of single-species individual-based model (ESIBM) / Spatial / Dynamic / Individual-based / 1 day or smaller / (1) One focal species (explicitly considered);
(2) One or several prey species (implicitly considered; fields of biomass for these species are used to force the model). / Growth, survival, reproduction, movement / No calibration; only some parameter tweaking.
MSVPA (multispecies virtual population analysis) and MSFOR (multispecies forward simulation) / Non-spatial / Dynamic / Age-structured / 1 year / Several species, including one focal fish species and several other species (e.g., fish and marine mammal species), which are either predators or prey or the focal species. / Growth (possibly), survival, reproduction (possibly) / Typically, calibration using fisheries catches, age- and size-compositions from fisheries, survey indices, age- and size-compositions from surveys, and diet data from stomach samples.
Gadget (Globally applicable Area Disaggregated General Ecosystem Toolbox) / Spatial / Dynamic / Age- and size-structured / Typically 1 month / (1) Fish and invertebrate species (explicitly considered);
(2) Other species ("other food"; implicitly considered). / Growth, survival, reproduction, movement / Calibration using numerous time series (biomasses, fisheries catches, etc.).
MICE (Model of Intermediate Complexity for Ecosystem assessment) / Non-spatial or spatial, depending on the questions to be addressed / Dynamic / Trophodynamic or age-/size-/stage-structured or individual-based, depending on the questions to be addressed / 1 year or smaller time step, depending on the questions to be addressed / (1) A limited number of species and functional groups (explicitly considered);
(2) A limited number of other species and functional groups (implicitly considered). / Growth, survival, reproduction (possibly), movement (possibly) / No calibration or calibration using time series of biomasses and/or fisheries catches, depending on the questions to be addressed.
Energy flow model / Non-spatial / Steady-state / Trophodynamic / Depends on the application; the temporal resolution of an energy flow model can be as high as 0.1 day / Marine mammals; fishes; invertebrates; primary producers (phytoplankton groups, marine plants); nitrogen pools. / Growth, survival / No calibration, but rather "flow-balance” of the model.
Ecopath with Ecosim (EwE) with Ecospace / Non-spatial (Ecopath, and Ecosim)/ spatial (Ecospace) / Steady-state (Ecopath)/ dynamic (Ecosim, and Ecospace) / Trophodynamic; or trophodynamic and stage-structured (when some species/functional groups are modeled using multistanza life-history models) / 1 year (Ecopath)/ 1 month (Ecosim, and Ecospace) / Marine mammals; seabirds; sea turtles; fishes; invertebrates; primary producers (phytoplankton groups, marine plants); detritus; and, often, microbial groups. / Growth (through the use of multistanza life-history models; Ecopath, Ecosim, and Ecospace), reproduction, survival (through the use of multistanza life-history models; Ecopath, Ecosim, and Ecospace), movement (Ecospace) / Ecosim can be calibrated using time series of biomasses, catches and other data and a weighted sum of squared deviations procedure. The Ecosim calibration process traditionally involves tweaks in the "vulnerability" parameters, which allow Ecosim predictions to be in stronger agreement with observed time series.
CASM (Comprehensive Aquatic System Model) / Spatial (“pseudo-spatial”; see Table 1) / Dynamic / Trophodynamic / 1 day or smaller time step / Fishes; invertebrates; primary producers (phytoplankton groups, marine plants); detritus; bacteria. / Growth, survival, reproduction (in some applications) / The biomasses predicted by CASM are compared to observed biomasses; CASM parameters are then adjusted so that CASM predictions remain within the bounds of natural variability.
Atlantis / Spatial (3D) / Dynamic / Age-structured (vertebrate groups)/trophodynamic (invertebrate groups) / 1 day wherever possible; shorter time step for functional groups with a fast turnover, such as phytoplankton groups. / Nitrogen and silicon pools of living, dead, nutrient, physical components and gaseous components. The living components of an Atlantis model include: humans; marine mammals; seabirds; sea turtles; fishes; invertebrates, primary producers (phytoplankton groups, marine plants); detritus; bacteria; and sediment bacteria. / Growth, survival, reproduction, movement / Typically, Atlantis models are calibrated through an historical simulation using biomass time series; this historical simulation employs a number of drivers (e.g., historical catch trends, historical and seasonal/spatial fishing regulations, seasonal hydrodynamics, etc.).
NEMURO model (NEMURO.FISH and NEMUROMS.FISH) / Spatial (2D = horizontal dimension only; or 3D) or non-spatial (1D = vertical dimension only) (NEMURO.FISH)/ spatial (3D) (NEMUROMS.FISH) / Dynamic / Individual-based / Typically, 1 hour but sub-models operate at a higher temporal resolution. / Fish species; invertebrates (zooplankton), nutrients; and detritus. / Growth, survival, reproduction, movement / Calibration to time series of biomasses (NEMURO.FISH)/ calibration of fish growth, mortality, reproduction and movement, and the fishing fleet submodel, using simplified grids before insertion of the processes into the full end-to-end model (NEMUROMS.FISH).
OSMOSE (Object-oriented Simulator of Marine ecOSystems Exploitation) / Spatial / Steady-state / Individual-based / Typically, two weeks or 1 month / (1) Focal fish species/functional groups; and focal invertebrate species/functional groups (explicitly considered);
(2) Background fish species/functional groups; background invertebrate species/functional groups (including zooplankton groups); and phytoplankton groups (implicitly considered). / Growth, survival, reproduction, movement / Calibration to mean biomass data using a specific evolutionary algorithm.
InVitro / Spatial (3D). / Dynamic / Individual-based / Typically seconds. However, for computational reasons, InVitro submodels generally have a temporal resolution / The components of an InVitro model are all "agents"; these model components include: fish, crustaceans, sharks, turtles, benthic communities, seagrass and mangroves, fisheries, shipping, oil and gas production, salt extraction, coastal development, port maintenance (such as dredging), recreational activities (such as recreational fishing). Some of the agents are represented at the individual level, while others are aggregates. Functional and physical attributes are given to each of the agents, while rules are specified for movements, growth and mortality at appropriate spatio-temporal scales. The agents evaluate the local abiotic and biotic environment and respond appropriately. Agents in InVitro include (Gray et al., 2006): “(1) Animal agents, which are the predominant agents of InVitro. It (and its subordinates) is used to represent all mobile animals (e.g., prawns, finfish, sharks, and turtles). This high level structure draws heavily from the traditional individual based model structure. It deals with the movement, mortality and basic behavior of animal life. Derived agents deal with variants or elaborations of this structure. The derived agent types also deal with specific life history characteristics, such as air breathing, flight, and spawning of larval stages. The predation of animal agents depends, among other things, on gape size and the predators gut capacity parameter (proportional to body size). (2) Population agents, which are a subordinate form of the agent type. Their formulation is a variant of the age-structured models used widely in fisheries. Rather than represent the entire population within a single agent, a number of Population agents are used, each representing a sub-population with its own center of gravity, movement, behavior and history. (3) Polyorganism agents, which are used to represent flora and fauna best modeled as patches (2D or 3D) rather than individuals or schools (e.g., patches of larvae or plankton, mangrove forests, seagrass meadows, stands of macroalgae and reefs). The decision structure behind this agent type is similar to that for the mobile agents, but incorporates more features typical of standard metapopulation models. The most effective means of representing most polyorganism agents is via polygons. Individual polygons within a set of polygons have their own attribute values and represent local dynamics of a wider population.
(4) Benthic agents, which are subclasses of the Polyorganism agents (e.g., habitat defining groups, such as seagrasses and mangroves). They are represented using a metapopulation model framework that tracks the evolution of percentage cover through time. (5) Blastula agents, which are another subclass of polyorganism agents. These are the primary means of representing juveniles that live separately from the adult population. (6) Larva agents, which are another subclass of polyorganism agents and are an alternative to Blastula. They undergo specific behaviors, such as advection-diffusion, directed movement to settlement sites, juvenile growth, and directed movement to adult sites. (7) Adviser agents, which are used either as an overall spatial record keeper, where every taxon in the model is recorded in the same grid for easy visualization of overall biomass patterns when considering model outputs, or as a general habitat and environmental conditions monitor. (8) Catastrophic agents, which represent two types of catastrophic events: natural events such as cyclones, and anthropogenic actions (e.g., dredging). The catastrophic events that will occur in a simulation run are read-in from a file at the beginning of the run (which details their timing, spatial location and intensity). (9) Contaminant agents. (10) Population biomass and fish biomass agents. (11) Vessel agents, which are used to simulate the freighter and tanker traffic that traverse the marine environment. (12) Port and fixture agents, which do not have many specific features beyond being locations or markers that are used as destinations or navigation beacons by other agents. They typically represent travel route waypoints, markers, buoys, and the source of the contaminant plumes. (13) Boat agents, which are a type of Vessel agent. They are responsible for fishing. (14) Plane and trap agents, which have a supporting role in the fisheries activities of the Boat agents. (15) Recfisher agents, which represent the recreational fishing pressure imposed by the local human population on the Population and Animal agents. (16) Fisheries Management Authority (FMA) agents, which are a form of management agents. They assume that the fisheries are being modeled and managed using a combination of reference points and gear, effort or spatial and temporal closures. The typical form of the FMA agents uses a fisheries stock assessment model and the decision procedure that uses them. (17) Department of Transport (DOT) agents, which monitor the arrivals and departures of transport vessels to the major ports in the modeled area, as well as the number of times these vessels have to invoke evasive maneuvers in order to prevent a collision while in coastal waters, as an indicator of a vessel collision and possible spill. (18) Environmental protection agency (EPA) agents, which are used to control the level of contaminants permitted at outfalls, and to trigger contaminant related management actions decreed by other management agents (such as the FMA agents). (19) Purity agents, which check for the probity of each agent of InVitro.” / Growth, survival, reproduction, movement. / InVitro models are calibrated to historical commercial catch-per-unit data.
Size spectrum model / Non-spatial / Dynamic / Size-structured / 1 year / (1) Several fish species;
(2) Unstructured resource spectrum, which represents plankton and simply serves as food for the smallest individuals of the community explicitly considered in the size spectrum model / Growth, survival, reproduction / No calibration; only some parameter tweaking.
Table OR3.2. Other important features of existing ecosystem modeling approaches
Ecosystem modeling approach / Other important features / Reference publications /Conceptual model / In conceptual models, the status of ecosystem services trigger responses (management actions) that have an impact on pressures and/or states and/or services of the ecosystem of interest. / Kelble et al. (2013)
Loop analysis / (1) A model can be represented as a signed directed graph, where variables (individual species and/or functional groups) are depicted as nodes. The connection between nodes is made through links, which represent the sign of direct effects. An arrow (→) represents a positive interaction and a line with a filled circle (—•) signifies a negative effect, while lines originating and terminating at the same node signify self-effects. (2) A set of Lotka–Volterra equations can also be employed to describe interacting populations: dNi / Ni*dt = A*Ni + k; the left side of the equation represents the rate of change in the abundance of a population i, which is controlled by birth, death and migration rates. Birth, death and migration rates can be embodied as interactions within and between populations (summarized in the community matrix A) or effects intrinsic to the population (summarized in the vector k). Each element of the community matrix A, aij, represents the direct effect of a variable j on a variable i. Therefore, the sign structure of the community matrix A is equal to the sign directed graph of the modeled system. (3) The resistance of the modeled system to a long-term perturbation can be evaluated by calculating the determinant of the community matrix A, whereas the relative response of the different variables to a perturbation can be assessed by computing the adjoint matrix of the community matrix. / Dambacher et al. (2003); Marzloff et al. (2011)