There are numerous ways to develop an Ecospace model, all of which have in common that the first step should always be to ensure that a spatially-explicit ecosystem models is actually required to evaluate the research questions at hand. With that being the case, the same three-step process to get started is

1. develop and balance an Ecopath model;
2. calibrate the balanced model in Ecosim; and
3. define the spatial model domain in Ecospace.

The core of the Ecospace workflow begins with a balanced Ecopath model. Thus, once the research questions have been clearly defined, the first step is to develop (or adopt) an Ecopath model. See Heymans et al. ^[1] for the best practices in developing a balanced Ecopath model and Ainsworth and Walters ^[2] for the most common mistakes to avoid when using EwE. After the Ecopath model is balanced, the best practice is to use the Ecosim temporal simulations and its calibration capabilities as the ‘go-to’ calibration method prior to building an Ecospace model. This approach allows one to assess the capabilities of the model to represent past dynamics under past forcing conditions (hind-cast calibration) and assess the impact of the strength of the prey-predator interactions (or vulnerability parameters) prior to expanding to the more complex Ecospace module. This also helps provide an understanding for the main temporal drivers of the system and their impacts. The current and primary benefit of using Ecosim to calibrate the model is that calibration has yet to be included in Ecospace. There are more informal ways to ‘visually’ calibrate the spatial model (see Spatial Model Skill Assessment), but there has yet to be a tool developed within the software to systematically calibrate an Ecospace model ^[3].

Many recent examples show the implementation of the Ecosim module of EwE for calibration of time-dynamic models (e.g., ^{[4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]}).

The general method for calibration using the Fit to Time Series module in Ecosim and detailed methods using this tool have been well-documented ^{[15] [16]}. However, when a traditional fit cannot be achieved due to lack of time series data, exploratory analyses to determine the uncertainty associated with changes in vulnerabilities is recommended (e.g., Rehren et al. ^[17]).

The primary goal of fitting the Ecosim model to time series data is to evaluate if the model can reproduce historical patterns while refining key parameters, such as the vulnerabilities ^[18]. Ecosim simulations are especially sensitive to the ‘vulnerability’ settings, which incorporates density-dependence by representing behavioral ecology responses by prey that can limit predation impacts and access to prey food resources, and expresses how far a group is from carrying capacity ^[19].

Fitting a model to data is a process that usually involves an iterative approach, using long-term fisheries, biological, and environmental data together with vulnerability settings to evaluate which combination of data and vulnerabilities minimizes residuals relative to observations calculated during the fitting procedure. This calibration process also allow for evaluation of model sensitivity to forcing functions and vulnerability setting, and can provide a priori information on how the user should conduct a model sensitivity analysis (usually using the built-in Monte Carlo routine plugin).

The data to incorporate into the calibration process cannot be predefined, and ultimately depends on data availability for the modelled ecosystem. A general rule of thumb is to have as long a reference period as possible to capture historical disturbance patterns and productivity shifts that may have occurred through both natural and anthropogenic fluctuations. Moreover, we can reflect data quality and error in the model fitting process. We can weigh the contributions made by different species to the goodness of fit term differently to reflect data quality. After defining what observational data to use for calibration, the sum of squared model deviations (SS) and Akaike Information Criterion (AIC) values should be used to find the most parsimonious model configuration that captures the most variation of the historical period in consideration ^[20] [21]. Calibration in Ecosim is not required for running Ecospace, but if done, the model is more reliable and interpretable, and more likely to lead to temporal dynamics in Ecospace for most species and functional groups, which are similar to the underlying Ecosim model.

Adaption

The chapter is in part adapted, with permission,  from:

De Mutsert K, Marta Coll, Jeroen Steenbeek, Cameron Ainsworth, Joe Buszowski, David Chagaris, Villy Christensen, Sheila J.J. Heymans, Kristy A. Lewis, Simone Libralato, Greig Oldford, Chiara Piroddi, Giovanni Romagnoni, Natalia Serpetti, Michael Spence, Carl Walters. 2023. Advances in spatial-temporal coastal and marine ecosystem modeling using Ecopath with Ecosim and Ecospace. Treatise on Estuarine and Coastal Science, 2nd Edition. Elsevier.