Ecospace limitations (edit)

While versatile, Ecospace has a number of conceptual and operational limitations. Conceptual limitations are related to the limits of the software. One conceptual limit stems from the fact that the spatially explicit model (Ecospace) is an expansion in space of a model that is not spatially explicit (Ecopath with Ecosim). The additional parameters and the initial conditions in Ecospace have a large influence on the spatial-temporal dynamics. The biomasses and catches, therefore, can be considerably different from Ecosim (Coll et al., 2016; Püts et al., 2020). In cases where spatial overlap of predator and prey is insufficiently to meet dietary needs, and where fishing effort or species foraging capacity are constrained to relatively small areas of the map, Ecospace might simulate over-depletion of the target species, which might not arise in a temporal (only) framework (Ecosim).

A second conceptual limit is that Ecospace is not three-dimensional: the water column cannot be divided into explicit layers. Ways to overcome this limitation include considering depth implicitly, through the inclusion of drivers of depth, or of environmental drivers such as using surface and bottom temperature. An alternative strategy is to separate benthic and pelagic components of the system (i.e., groups), so that trophic groups are not available to one another (i.e., no or limited diet overlap), even if they are physically distributed in the same 2D area. This approach requires careful consideration of the species groupings, highlighting that planning of a model that is to be used in Ecospace should consider this from the start, at the model topology stage.

Another conceptual limit is the fact that in Ecospace fishing effort is distributed over the map (excluding those closed to fishing) based on a gravity model, that accounts for profitability, which considers the catch portfolio, price, as well as fixed and variable (e.g. sailing) fishing cost (Walters et al., 1999). This causes the model to predict effort in areas that are most profitable. However, factors other than profitability influence effort distribution such as avoidance for by-catch species, fidelity to location, and expected or perceived economic revenues (Cabral et al., 2017; Collins et al., 2021; Poos et al., 2010; Poos and Rijnsdorp, 2007; van Putten et al., 2012). One workaround is to use preference as a weight factor of price, i.e., quantifying the fishing preference for a determinate species, and use this to drive the effort distribution. Another option could be to set the sailing cost map based on the (inverse of) existing effort patterns, rather than by distance to port, which if invoked is the default. In this way, the model will predict lower cost in areas where the fleets are historically observed and predict higher effort in these areas.

Operational limits relate to the possible issues encountered in the parameterization and setup of an Ecospace model. For example, multi-stanza groups must be used with caution, and need to be carefully assessed. The spatial displacement between juvenile and adult stanzas with ontogenetic diet shifts can lead to an unrealistic increase or decrease in juveniles (if insufficient/excessive predation is present in the respective areas), leading to oscillatory patterns that may cascade through the system. These patterns need to be adjusted through careful consideration of vulnerability and diet inputs.

Spatial resolution matters and the decision about cell size should be based on what type of questions the model will be used for, although most commonly it is driven by data availability. Higher resolution has computational demand and is not necessarily beneficial for all analyses. Coarse resolution, on the other hand, leads to simplification of topology, habitats, MPAs, and other features. Unexpected behavior can arise; instead of an intense but localized impact (e.g., noise, harmful algal blooms), Ecospace will smooth an impact across a few cells. This might be unsuitable for some applications: a preliminary analysis is needed to assess if the resolution can appropriately capture the expected scale of the impacts to be studied. Resolution also affects the representation of patchy habitats: for example, isolated cells and long and narrow strips of cells of a certain habitat can cause the biomass to get stuck, creating unrealistic patterns or biomass crashes. Finer resolution helps to reduce such issues. Specific plug-ins to overcome challenges related to spatial resolution issues have been implemented in the last years, such as the Biomass Emitter plug-in (Steenbeek et al., 2018; see 6.3) and the EcoEngineer plug-in (Steenbeek and Sadchatheeswaran, 2021; see 6.9).

The temporal resolution is by default limited to monthly time steps; however, specific processes such as larval dispersal followed by settlement or tidal movement are not well captured at monthly time steps. While the advection module can help to capture larval dynamics, these are still driven by monthly dynamics that may or may not be resolved enough to capture spatiotemporal larval distribution realistically. The Ecospace IBM model moves stanza packets at sub-monthly time steps. In theory, as all units including time are only scaling factors within the EwE approach, input data can be scaled to daily averages, which will implicitly run EwE in daily time steps (Libralato and Solidoro, 2009). Note that this might cause differences in Ecosim results because of the integration at shorter time step and problems with the Ecospace advection logic and all multi-stanza logic that has hard-coded cycles of twelve based on annual patterns embedded within. When rescaling time units, this limitation needs to be kept in mind.

As an example, the IBM model of Ecospace was used to model tidal dispersal of out-migrating salmon smolt in an estuary by one of the co-authors (unpublished). For this a model time-step of 12 hours was used so that rates were defined in the Ecopath model with the unit of per 12 hours, and each time step for runs in Ecospace was one hour allowing for modeling the impact of the tidal cycles.

Ecopath biomasses are initially distributed over the spatial domain in accordance with proportions of the habitat forage use and response to environmental drivers. Fisheries effort is initially distributed according to biomass of the target species. These initial conditions, even in the absence of time varying environmental drivers, need to be adjusted and Ecospace biomass takes some simulated years before they can reach equilibrium. The length of this adjusting phase depends on the complexity of habitat definition and preferences, and on dispersal rate parameters and should not be confused with forced spatial dynamics. Therefore, it is good practice to set a burn-in period with no forcing or disturbances, i.e., the spin-up preceding the perturbed simulation. The number of spin-up years can be determined by running the model first without forcing functions and identifying the time needed to reach equilibrium, i.e. stable state. Notably the spin-up should also include no changes in Ecosim, in order to have steady distribution of species only due to spatial adjustments in Ecospace. Thereafter the model can be run with forcings, included after the spin-up years, and it can be assessed whether the spinoff period is robust enough. Forcing varying in time aiming to replicate spatiotemporal patterns should start after the spin-up time, as the model base year can be assumed to correspond to the end of the spin-up period. A build-in spin up period capability is available in the software under a professional license.

Some of the limitations that have emerged in Ecosim have been gradually addressed through continuous model advancements: for example, explicit inclusion of uncertainty, or refinement of the fitting to time series with statistical advancements. In Ecospace, some of these aspects are still under development: at present, spatial-temporal fitting is under development, and uncertainty in parameters cannot be fully accounted for. Similarly, fisheries management tools that are well developed in Ecosim are not as advanced in Ecospace (e.g. Marine Strategy Evaluation, Optimization Routine, Maximum Sustainable Catch search, etc.). However, beyond the computational and financial issues, there are no substantial limitations for these extensions, some of which are already underway.